Behavioral Response Analysis of Rural Residents’ Living Waste Classification: Evidence from Jiangsu, China

Abstract

Maximizing rural residents’ primary role in domestic waste sorting and management is critical to improving the rural living environment and advancing socioeconomic development. This study aims to analyze the entire process of domestic waste sorting by rural residents using sample data from 2420 rural households surveyed in the 2021 China Land Economic Survey (CLES). Based on the Theory of Planned Behavior (TPB), this study develops a research framework for analyzing the entire process of rural residents’ domestic waste-sorting behavior. It examines the inter-relationships among behavioral cognition, behavioral intention, and behavioral response and employs structural equation modeling (SEM) for empirical verification. The results demonstrate that subjective norms, classification attitudes, and perceived behavioral control exert statistically significant positive effects on both rural residents’ intention and behavioral responses toward domestic waste sorting. Moreover, sorting intention demonstrates a significant predictive effect on actual sorting behavior. This study further identifies a mediating role of sorting intention throughout the behavioral process, while potential correlations among subjective norm, behavioral attitude, and perceived behavioral control suggest additional mechanisms underlying rural residents’ waste-sorting responses that warrant further exploration.

1. Introduction

As China’s rural revitalization strategy continues to deepen, the rural habitat environment has significantly improved; however, the management of rural household waste persists as a critical challenge [1]. China’s rural domestic waste generation has expanded at an average annual rate of 6.8% over the past decade. According to a recent survey by China’s health department, the current per capita daily production of rural living garbage in China stands at approximately 0.86 kg. Improper disposal of rural domestic waste remains a critical factor contributing to environmental degradation in rural residential areas [2] and also hampers agricultural development [3]. International experience indicates that source separation of domestic waste is a vital method for effective waste management [4,5], reducing disposal costs and enhancing resource utilization [6]. In China’s rural context, residents themselves are the principal agents of source classification, with their active participation profoundly influencing the efficiency of waste sorting, collection, transfer, and treatment.

Despite these obvious benefits, rural household waste classification still faces substantial obstacles. Although scholars have examined various factors affecting residents’ willingness to participate [7], research focusing on urban contexts predominates [8], leaving rural waste classification relatively underexplored [9,10]. In particular, existing studies often overlook the complete behavioral chain from farmers’ environmental awareness to their ultimate classification actions [5,11]. This gap is problematic because while some residents acknowledge the importance of proper waste management, the limited direct benefits and perceived complexities of sorting often reduce their actual participation [9]. Such inconsistencies underscore the need to examine subjective norms, attitudes, and perceived behavioral control—critical constructs in environmental behavior research—within the rural setting. Without translating awareness into actionable behaviors, research on rural waste sorting loses its practical significance.

In an effort to encourage rural waste source separation, both central and local governments have introduced and implemented targeted policies and regulations. The Law of the People’s Republic of China on Promoting the Revitalization of Rural Areas stipulates promoting “simple and locally adaptable” garbage separation. Central Government Document No. 1 (2025) likewise calls for improving domestic waste separation in rural areas, while the “Five-Year Action Program for Improving and Upgrading Rural Habitat (2021–2025)” suggests exploring simple, farmer-friendly classification approaches. Nevertheless, participation from the principal governance agents—village committees, government departments, and especially farmers—remains insufficient [12]. The Ministry of Agriculture and Rural Affairs surveys reveal that only 35% of rural households can accurately distinguish between recyclable and hazardous waste, while actual active sorting rates remain below 20%. The Ministry of Ecology and Environment inspections in 2023 further found that 62% of villages maintain waste-sorting policies at the promotional stage, lacking substantive monitoring mechanisms. Consequently, a comprehensive examination of rural residents’ responses to household waste classification is essential for identifying the factors that most effectively promote sorting behavior.

Against this backdrop, this study constructs a full-process cognition–intention–behavior framework grounded in the Theory of Planned Behavior (TPB) using structural equation modeling (SEM) to investigate determinants of rural residents’ waste-sorting intentions and behavioral responses. Such analysis is pivotal for guiding robust policy formulation, designing targeted interventions, and establishing a viable institutional framework for rural domestic waste management.

2. Literature Review

Research on waste classification has a longstanding history, initially emphasizing policy orientation, material flow assessment, the valuation of classification behaviors, and the identification of key factors influencing waste-sorting practices [13,14,15]. With the dawn of the 21st century, international studies on waste classification expanded significantly, driving the development of comprehensive indicator systems and theoretical models for predicting sorting behaviors. The scope of research has broadened from individual-level analyses to community, urban, and even global scales.

Prominent theoretical frameworks include the Theory of Planned Behavior (TPB), the Information–Motivation–Behavior (IBM) model, the Attitude–Behavior–Circumstance (ABC) model, the Social Capital Theory, and the Multicenter Governance Theory [16,17,18]. Among these, Multicenter Governance Theory particularly draws from psychology, economics, and sociology to examine how individuals and groups engage in waste-sorting practices. By analyzing spatial differences in public participation under varying regional, hierarchical, and contextual factors, as well as how these factors interact at multiple levels, this approach offers policy recommendations for both urban and rural waste management. Rural areas lag far behind urban centers in infrastructure, and farmers face greater challenges in acquiring waste-sorting knowledge compared to urban residents. Consequently, governments implement differentiated waste-sorting policies and models based on urban–rural disparities, leading to distinct urban–rural disparities in waste-sorting research [19]. Owing to its robust explanatory power for rural social norms and its comprehensive cognition–intention–behavior framework, the Theory of Planned Behavior (TPB) emerges as the optimal theoretical tool for analyzing the full-process responses to rural waste sorting. Conversely, the Information–Motivation–Behavior (IBM) and Attitude–Behavior–Context (ABC) models exhibit theoretical gaps in accounting for rural social complexities and behavioral constraints, thereby limiting their capacity to inform precision intervention strategies.

Recently, in alignment with China’s rural revitalization strategy and new policies promoting rural habitat improvement, scholars in China have intensified their exploration of rural household waste classification. The existing literature has focused on diverse themes, including the design of classification modes, the governance challenges inherent in waste-sorting programs, farmers’ willingness to sort waste, the behavioral processes underlying classification, and gaps between individuals’ intentions and their actual sorting behaviors [20,21,22]. Empirical findings suggest that farmers’ willingness to participate in waste classification hinges on a combination of individual characteristics—such as income level, social capital, length of village residency, and environmental awareness [23]—and broader socioeconomic factors, including recycling rates, social supervision, institutional trust, and environmental concerns [24]. Existing studies commonly indicate that analyzing farmers’ classification intention and behavior can effectively predict the effectiveness of household waste source classification. Based on this, a multi-dimensional indicator system covering classification intention and behavior has been developed to measure rural waste classification performance. Key indicators include the presence of classification intention [18], the strength of classification intention, participation in classification, implementation level of classification behavior, frequency of classification behavior, and sustainability of classification behavior [25].

Despite the richness of these studies, crucial research gaps remain. First, past scholarship has often focused on urban contexts, leaving rural waste classification relatively underexplored. Given that rural residents are both the producers and the central agents of waste management, understanding their behaviors, motivations, and constraints is essential yet insufficiently addressed. Second, many existing investigations are limited to discussing either farmers’ willingness to classify or the outcome of sorting behavior without systematically analyzing the entire cognitive–behavioral process. Such an approach overlooks the mediating roles of subjective norms, attitudes, and perceived behavioral control, as well as the possible discrepancy between the willingness to act and actual behavior.

Therefore, a deeper exploration of the full decision-making chain—from farmers’ cognition (awareness, attitudes, and perceived control) to their intention and ultimate behavior—can offer more robust theoretical insights and practical recommendations. Filling this gap is particularly vital for rural areas in China, where household waste classification systems remain in nascent stages. By comprehensively analyzing the factors shaping farmers’ sorting responses and situating these within the broader policy and sociocultural context, this study aims to inform the design of targeted interventions and improve the efficacy of rural waste management initiatives.

The aforementioned studies have established a foundation for analyzing waste classification behavior among rural residents; however, opportunities for further research remain. First, while existing research predominantly focuses on the factors influencing waste classification behavior, it less frequently addresses the comprehensive analysis of the behavioral response process. Second, the scope of research is narrow, primarily concentrating on urban settings, with a limited focus on the behaviors of rural residents in China. Moreover, the minimal participation of the principal governance agents in rural areas currently hinders improvements in the rural living environment, with existing studies insufficiently examining the role of farmers’ behavior [18]. Therefore, constructing a full-process cognition–intention–behavior framework to analyze the entire process of behavioral responses to domestic waste sorting among rural residents in China and explore the effective determinants of these behaviors is essential.

This paper offers significant contributions in terms of research perspective, methodology, and content. Firstly, it explores the entire process of rural residents’ domestic waste classification behavior and examines the factors influencing their behavioral responses from the perspective of farmers. Secondly, it incorporates the Theory of Planned Behavior into the study of rural residents’ waste classification responses and employs structural equation modeling for analysis. Lastly, it examines factors influencing rural residents’ willingness to classify and their behaviors, offering policy recommendations to enhance their participation in domestic waste management.

3. Theoretical Framework

3.1. Theoretical Foundation

The Theory of Planned Behavior (TPB), proposed by Ajzen, is a behavioral decision-making model designed to predict and understand individual actions [26]. This theory explains how individuals plan and execute behaviors to achieve specific goals. Widely applied across diverse domains, TPB has been extensively used to analyze and predict behavioral decisions in health-related behaviors [27], entrepreneurship [28], fertility choices [29], and pro-environmental actions [30]. The Theory of Planned Behavior posits that an individual’s behavioral intention, which is directly influenced by three latent variables—behavioral attitudes, subjective norms, and perceived behavioral control—determines behavioral responses. These variables concurrently affect the implementation of the behavior. This theory elucidates the mechanism and rationale behind behavior generation by examining the logical relationships among behavioral attitudes, intentions, and responses [26].

Rural residents’ domestic waste classification behavior is fundamentally a form of planned decision making [31]. Accordingly, this study constructs a model under the Theory of Planned Behavior, tracing the pathway from cognitive judgment to intentional choice to behavioral response (Figure 1). The model comprises two components: rural residents’ cognition related to domestic waste classification and their behavioral response to this classification. The cognitive component encompasses rural residents’ behavioral attitudes, subjective norms, and perceived behavioral control regarding waste classification. Given that behavioral intentions dictate responses when control conditions are adequate, the behavioral component consists of the residents’ willingness and actual behaviors toward waste classification. Drawing on these theories, grounded on sound assumptions, this study investigates the mechanisms of rural residents’ domestic waste classification behaviors [22].

3.2. Hypothesis Development

Subjective norms are the sum of normative beliefs and social pressure compliance motives, reflecting the degree to which an individual’s willingness to adopt a particular behavior is influenced by other relevant and important people or groups [32]. In this study, subjective norms reflect the social pressure perceived by rural residents regarding waste sorting. Their waste-sorting behavior is influenced by family, friends, and neighbors. If those around them actively participate in waste management and exert peer supervision, it enhances residents’ own willingness to engage in proper waste sorting [33]. In terms of the behavior of living garbage classification of rural residents, it is necessary for many stakeholders, including government departments, village committees, and the majority of farmers, to participate in cooperation and promotion. In this collective action, external claims and pressures are bound to promote or constrain individual willingness and their behavioral response [34,35]. Social norms and interpersonal norms represent the degree of pressure on residents to adopt waste separation behavior. From the perspective of social norms, the village committee’s response to rural residents’ garbage sorting behavior affects the degree of villagers’ response to living garbage sorting, so whether villagers perceive that living garbage sorting is appreciated and praised and whether villagers are satisfied with the ecological livability of their village directly affects the willingness of farmers and their degree of behavioral response [31]. From the perspective of interpersonal norms, the vast number of villagers in the village are closely connected to each other and influence each other, so the villagers’ satisfaction with the classification of domestic garbage in the village and the villagers’ attitudes toward the environmental protection behaviors of other farmers are also key factors [18]. Accordingly, the following hypotheses are proposed: H1, subjective norms have a positive and significant effect on rural residents’ willingness to classify household garbage; H2, subjective norms have a positive and significant effect on rural residents’ behavioral response to classify household garbage.

Behavioral attitudes are a combination of behavioral beliefs and outcome assessments, which are formed by individuals based on their favorable or unfavorable judgments of the possible consequences of specific behaviors [32,33]. It has been shown that attitude is the strongest predictor of willingness [36], and positively reinforcing the attitude of residents’ garbage classification behavior can stimulate the enthusiasm of rural residents’ living garbage classification management and influence the willingness and behavioral response of rural residents’ garbage classification [24]. In rural residents’ waste-sorting behavior, as “rational economic agents”, they form behavioral beliefs and outcome evaluations based on expected benefits, which shape their willingness to participate. In rural waste-sorting initiatives, behavioral attitudes derived from expected benefits are primarily measured in terms of ecological and social responsibility dimensions [17]. First, whether rural garbage classification plays a positive role in environmental improvement and the degree of impact of unclassified household garbage on the community environment and the order of life in the village reflect the villagers’ attitudes toward household garbage classification from the perspective of ecological and environmental benefits; second, the villagers’ sense of responsibility to carry out household garbage classification reflects the villagers’ attitudes toward household garbage classification from the perspective of social responsibility [37]. Theoretically, the villagers’ good knowledge of the expected benefits of domestic garbage classification will promote the villagers’ willingness to participate and the positive degree of action response [38]. Accordingly, the following hypotheses are proposed: H3, behavioral attitudes have a positive and significant effect on rural residents’ willingness to classify household garbage; H4, behavioral attitudes have a positive and significant effect on rural residents’ behavioral response to classify household garbage.

Perceived behavioral control is a combination of perceptual power and control beliefs, reflecting the individual’s perception of the difficulty of adopting the actual behavior [32,33]. Generally speaking, the more rural residents know about environmental protection and the more they know about the methods and measures of garbage classification and management, the higher their willingness to classify and manage household garbage and the higher the likelihood of adopting the behavior of classifying household garbage [39]. Moreover, the improvement in classification conditions by relevant government departments and the provision of a platform for residents to obtain information on garbage classification can increase the motivation of residents to manage household garbage and increase their willingness to classify. Perceived behavioral control can be measured in terms of control factors and control beliefs [40]. In terms of control factors, the promotion of living garbage classification in rural areas involves many subjects, is highly specialized, and has a long cycle. Villagers can fully perceive the difficulty, complexity, and uncertainty of the promotion of garbage classification; therefore, whether villagers know about living garbage classification and whether the government publicizes the classification of living garbage in rural areas will inevitably affect their willingness to classify and their behavioral response. In terms of control beliefs, the greater the expected benefits of garbage classification, the greater the control beliefs of the villagers and the greater their willingness to participate and act positively [41]. Therefore, whether or not there are material rewards for domestic garbage classification will affect their behavioral control beliefs. Accordingly, the following hypotheses are proposed: H5, perceptual behavioral control has a positive and significant effect on rural residents’ willingness to classify household garbage; H6, perceptual behavioral control has a positive and significant effect on rural residents’ behavioral response to classifying household garbage.

Behavioral willingness expresses “the tendency to try to carry out a certain behavior”, and behavioral response is the behavior in which individuals actually take action [32,33]. Willingness to sort refers to the time, effort, and resources an individual is prepared to devote to waste sorting. Among rural residents, this willingness is directly influenced by behavioral attitudes, subjective norms, and perceived behavioral control while simultaneously exerting a direct impact on sorting behavior. Other behavioral factors may also indirectly affect sorting behavior through their influence on sorting willingness [42]. In the promotion of rural residents’ garbage classification behavior, villagers’ behavioral willingness and intensity can reflect the strength of farmers’ behavioral willingness. Therefore, whether villagers have the will to carry out garbage classification can be used to measure the villagers’ own willingness, and whether they are willing to pay for the garbage classification work and how much they pay can be used to measure the strength of villagers’ willingness. Whether villagers carry out household garbage classification, how they deal with it, and the number of times they supervise others carrying out garbage classification can all reflect the degree of the response to villagers’ garbage classification behavior. Theoretically, the stronger the villagers’ willingness to participate in the separation of domestic waste, the higher the level of response to the action [43]. Meanwhile, the villagers’ cognition of the three dimensions indirectly affects the actual behavior of farmers by influencing their willingness to act. Accordingly, the following hypotheses are proposed: H7, behavioral willingness has a positive and significant effect on the behavioral response of rural residents’ living garbage classification of farm households; H8, in terms of the behavior of rural residents’ living garbage classification, the behavioral willingness of farm households plays a mediating role between cognition and the behavioral response.

In conclusion, this paper proposes the following hypotheses:

H1. 

Subjective norms have a positive and significant effect on rural residents’ willingness to separate household waste behavior.

H2. 

Subjective norms have a positive and significant effect on behavioral responses to rural residents’ household waste separation.

H3. 

Behavioral attitudes have a positive and significant effect on rural residents’ behavioral intention to separate household waste.

H4. 

Behavioral attitudes have a positive and significant effect on the behavioral response of rural residents to sorting household waste.

H5. 

Perceived behavioral control has a positive and significant effect on rural residents’ behavioral intention to separate household waste.

H6. 

Perceived behavioral control has a positive and significant effect on rural residents’ behavioral response to household waste sorting.

H7. 

Behavioral intention has a positive and significant effect on the behavioral response of rural households in sorting domestic waste for rural residents.

H8. 

Farmers’ behavioral willingness mediates between cognition and behavioral response in rural residents’ household waste-sorting behavior.

4. Materials and Methods

4.1. Data Sources

This study uses data from the China Land Economic Survey (CLES). CLES conducted its baseline survey in 2020 in Jiangsu Province, followed by follow-up surveys in 2021 and 2022. To select the sample counties and administrative villages, the survey employed the Probability-Proportional-to-Size (PPS) sampling method—a technique designed to ensure that larger units (in terms of population or other size measures) have a proportionally higher likelihood of being selected. Concretely, the survey team first determined a list of potential counties, then drew sample counties, townships, and villages with probabilities corresponding to their respective population sizes. This approach helps minimize sampling bias by reflecting the actual distribution of the population more accurately than conventional random sampling.

In total, 26 counties (or districts), 52 townships, and 52 administrative villages were sampled from the 13 prefecture-level cities in Jiangsu Province, yielding 2600 rural households as the initial sample. As the questions pertaining to rural household waste classification were systematically investigated in 2021, this study used the 2022 follow-up data for analysis. After data cleaning and removing incomplete responses, a final total of 2420 rural households were included in the study.

Table 1. Basic information of the sample.

Indicator	Definition	Proportion/Mean	Standard Deviation
Resident population	Number of people	3.07	1.60
Gender of respondents	1 = Male; 0 = Female	0.73	0.44
Age of respondents	One full year	62.26	11.46
Literacy of respondents	Number of years in school	7.12	4.21
Annual household income of respondents	Amount of money	30,313.84	54,161.17
Whether the respondent is a member of the Communist Party of China (CPC)	1 = yes; 0 = no	0.21	0.41
Health status of respondents	1 = incapacitated; 2 = poor; 3 = moderate; 4 = good; 5 = excellent	3.95	1.07

presents basic information regarding these sample households.

4.2. Selection of Variables

Based on the Theory of Planned Behavior (TPB) and related empirical findings, five latent variables—subjective norms (SNs), behavioral attitudes (BAs), perceived behavioral control (PEC), behavioral intention (BI), and behavioral response (BR)—were chosen to capture the full process of rural residents’ domestic waste classification. Each latent variable is measured by multiple observed indicators drawn from the survey.

Table 2. Variable selection and meaning.

Latent Variable	Item Label	Survey Question (Observed Variable)	Response Coing/Scale	Mean	SD
Subjective norms (SNs)	SN1	Do you think the separation of household waste can be appreciated and praised?	1 = Completely Disagree … 5 = Completely Agree	4.11	0.81
	SN2	What is your attitude towards other villagers’ environmental protection behaviors?	1 = Disagree … 3 = Strongly Agree (3-point scale)	2.44	0.55
	SN3	What is your satisfaction with the household waste separation in your village?	1 = Very Dissatisfied … 5 = Very Satisfied	3.55	0.96
	SN4	What is your satisfaction with the village’s ecological livability (village appearance, living convenience, sewage and garbage management, air quality, etc.)?	1 = Very Dissatisfied … 5 = Very Satisfied	4.16	0.77
Behavioral attitudes (BAs)	BA1	Do you agree that the separation of household waste has a positive effect on improving the rural environment?	1 = Completely Disagree … 5 = Completely Agree	4.25	0.91
	BA2	What do you think is the impact of haphazard piling/non-separation of household waste on the rural ecological environment?	1 = Very Small … 5 = Very Large	4.21	0.90
	BA3	What do you think is the impact of haphazard piling/non-separation of household waste on the community environment (village appearance and order of life)?	1 = Very Small … 5 = Very Large	4.24	0.84
	BA4	Do you think your household is responsible for sorting and placing household waste?	1 = Not Responsible at All … 5 = Fully Responsible	4.07	0.91
Perceived Behavioral Control (PEC)	PEC1	How well do you understand rural household waste separation?	1 = Have not heard of it … 5 = Know a great deal	3.25	1.21
	PEC2	1 = yes; 0 = no	1 = Yes; 0 = No	0.83	0.37
	PEC3	1 = disagree completely; 2 = do not quite agree; 3 = generally; 4 = fairly agree; 5 = completely agree	1 = Completely Disagree … 5 = Completely Agree	4.16	0.85
Behavioral Intentions (BI)	BI1	Are you willing to separate your household waste?	1 = Yes; 0 = No	0.97	0
	BI2	The process of rural household waste management sometimes requires a small fee. Would you be willing to pay?	1 = Yes; 0 = No	0.64	0.48
	BI3	If you are willing to pay, how many RMB per month would your household pay for rural household waste management?	(Open-Ended: amount in RMB	131.5	3237.53
Behavioral responses (BR)	BR1	Do you currently separate your household waste?	1 = Yes; 0 = No	0.53	0.50
	BR2	How is your household waste disposed of? 1 = not sorted, all garbage together 2 = two categories (recyclable vs. other) 3 = three categories (recyclable, food/putrescible, other) 4 = four categories (recyclable, food/putrescible, toxic/hazardous, other)	1 = No sorting … 4 = Four categories	1.73	0.87
	BR3	How many times have you supervised or reminded others (neighbors and family members) to sort garbage? 1 ≤ 3 times 2 = 3–5 times 3 = 6–10 times 4 ≥ 10 times	1 = Fewer than 3 times … 4 = More than 10 times	1.34	0.80

Notes: Latent variables: SNs = subjective norms, BAs = behavioral attitudes, PEC = perceived behavioral control, BI = behavioral intention, BR = behavioral response. Item labels: For clarity, each observed question under “Behavioral Attitudes” is labeled BA1, BA2, BA3, and BA4. You may likewise label other constructs’ items (SN1–SN4, PEC1–PEC3, etc.) if you wish to reference them similarly in the main text or results. Mean and SD: Means and standard deviations are shown for continuous or ordinal measures. For binary (1/0) variables, the mean can be interpreted as the proportion of respondents who answered “Yes”.

presents the revised variable labels and their descriptive statistics. To make referencing clearer, we labeled each question under behavioral attitudes (BAs) as BA1, BA2, BA3, and BA4 rather than listing them in a bulleted format (Table 2).

4.3. Model Specification

To systematically analyze the factors influencing rural residents’ domestic waste classification, we adopt structural equation modeling (SEM) [44]. SEM is a powerful multivariate statistical technique that allows us to model the latent constructs of interest—subjective norms (SNs), behavioral attitudes (BAs), perceived behavioral control (PEC), behavioral intention (BI), and behavioral response (BR)—and the structural pathways among these constructs [45,46]. This approach is especially suitable for testing complex theoretical frameworks such as the Theory of Planned Behavior in the domain of environmental behavior [5]:

4.3.1. Structural Equations

In matrix form, the SEM can be represented as follows:

$$\eta =\gamma \xi +\beta \eta +\zeta$$

(1)

$$Y=\lambda \eta +\epsilon$$

(2)

$$X=\lambda \xi +\delta$$

(3)

where the following apply:

$\eta =$ Endogenous latent variables (in our case, BI and BR).

$\xi =$ Exogenous latent variables (here, SNs, BAs, and PEC).

$\zeta =$ Residual term(s), capturing unexplained variance in the endogenous constructs. $Y$ and $X=$ vectors of observed indicators measuring $\eta$ and $\xi .$ ${\Lambda }_y$ and ${\Lambda }_x=$ factor loading matrices relating latent variables to their observed items.

In practice, $\eta$ is a vector of multiple endogenous variables, so the notation $\beta \eta$ does not imply that a single variable depends on itself. Instead, it means one endogenous variable (e.g., BR) may depend on another (e.g., Bl). A clearer representation is as follows:

$$\mathrm{B}\mathrm{I}={\gamma }_1\xi +{\zeta }_1,$$

(4)

$$\mathrm{B}\mathrm{R}={\gamma }_2\xi +{\beta }_{12} \mathrm{B}\mathrm{I}+{ \zeta }_2.$$

(5)

where BR depends on both the exogenous constructs, (ξ) and BI, rather than on itself.

4.3.2. Role of Latent Variables in Estimation

Our model treats SNs, BAs, PEC, BI, and BR as latent constructs, each measured using multiple survey items (e.g., SN1, SN2, BA1, BA2, etc.). These variables are not directly observed; they are inferred from their indicators. Using latent variables affects our estimation in several ways:

Measurement Error and Reliability

Because no single item can perfectly measure a complex concept like “behavioral attitude”, we specify a measurement model (Equations (2) and (3)) to confirm that each set of indicators indeed reflects its underlying latent factor.

This process partitions out measurement error, thereby improving the reliability and validity of our constructs compared to single-item or summated-scale approaches.

Two-Step SEM Approach

Step 1 (Confirmatory Factor Analysis, CFA): We first verify that each latent variable (SNs, BAs, PEC, BI, and BR) is measured well by its indicators. We assess factor loadings, composite reliability, and discriminant validity.

Step 2 (Structural Model): Once the measurement model is validated, we test the hypothesized paths among the latent variables (e.g., SNs → BI, BAs → BR, etc.). This step shows how well our theoretical model fits the data and how strongly each latent construct influences another.

Improved Parameter Estimates

By incorporating latent variables, SEM statistically adjusts for the random measurement error associated with each indicator.

The resulting path coefficients—like the effect of BAs on BI—are more robust and less biased than if we had treated each concept as a single observed indicator.

Model Fit and Indirect Effects

We gauge overall model quality using fit indices (e.g., RMSEA, CFI, TLI, and $x^{2/df}).$ Acceptable thresholds typically include RMSEA < 0.08 and CFI/TLI > 0.90.

SEM also enables testing mediating relationships. For instance, BI may mediate the effects of SNs, BAs, and PEC on BR. Without a latent variable framework, it is more challenging to verify these indirect effects with reduced bias.

Relevance to the Theory of Planned Behavior (TPB)

The TPB posits that attitude, subjective norms, and perceived control jointly shape intention, which then drives behavior.

Latent-variable SEM aligns naturally with TPB because each theoretical component (e.g., “attitude” or “subjective norms”) is conceptualized as a broad, multi-faceted construct.

In the context of rural waste classification, this comprehensive approach helps us identify which latent constructs have the strongest influence on residents’ willingness (BI) and actual classification behavior (BR), providing a more accurate basis for policymaking.

4.3.3. Estimation Procedure

We employ maximum likelihood (ML) estimation, which is common in SEM, assuming approximately multivariate normal data. If the initial model fit is suboptimal, we may introduce theoretically justified modifications, such as adding covariance paths between error terms of closely related indicators, to enhance model parsimony. We report conventional fit indices—RMSEA, CFI, TLI, NFI, RFI, etc.—to confirm whether the final model adequately explains the observed data.

In summary, handling SNs, BAs, PEC, BI, and BR as latent constructs within an SEM framework provides a rigorous, multi-faceted examination of rural waste classification behaviors. This approach not only mitigates measurement error but also facilitates the testing of direct and indirect effects in a single, integrated model. Consequently, the results yield stronger empirical support for targeted policy recommendations aimed at improving household waste-sorting behaviors in rural communities.

5. Analysis of Results

5.1. Reliability and Validity Tests

This study utilized SPSS 26.0 software to assess the reliability and validity of the scale, and the results are presented in the table below. Reliability evaluates the consistency of a scale in measuring the extent to which a questionnaire accurately reflects the actual situation. The commonly employed index for assessing reliability is the internal consistency based on standardized items, known as Cronbach’s alpha coefficient [47]. The test results indicate that the reliability coefficients for the dimensions in this study all exceed 0.8, demonstrating good content consistency across the measurement scales.

KMO, the Kaiser–Meyer–Olkin measure of sampling adequacy, indicates that higher values signify a greater presence of common factors among the variables, making it more suitable for factor analysis. According to Kaiser, the closer the KMO value approaches 1, the more appropriate the data are for factor analysis; values below 0.5 indicate reduced suitability [48]. The results reveal that the KMO values for the service quality items—0.821, 0.719, 0.729, 0.73, and 0.842—indicate a significant presence of common factors among the variables, making them highly suitable for factor analysis. Additionally, the χ² values from Bartlett’s test of sphericity—2008.897, 1295.534, 1465.447, 1596.189, and 2708.109—achieved significant levels (p < 0.001), indicating the presence of common factors within the correlation matrices, thereby confirming their suitability for factor analysis (

Table 3. Variable reliability and factor analysis results.

Impact Pathways	Standardized Factor Loading	C. R-Value	Cronbach’s α	KMO Value	Bartlett’s Test of Sphericity
SN4 ← SNs	0.708	24.485 ***	0.853	0.821	2008.897 (p = 0.000)
SN3 ← SNs	0.811	28.232 ***
SN2 ← SNs	0.768	26.471 ***
SN1 ← SNs	0.791
PEC3 ← PEC	0.823	24.934 ***	0.824	0.719	1295.534 (p = 0.000)
PEC2 ← PEC	0.757	24.225 ***
PEC1 ← PEC	0.761
BR1 ← BR	0.788		0.844	0.729	1465.447 (p = 0.000)
BR2 ← BR	0.808	27.331 ***
BR3 ← BR	0.819	27.675 ***
BI1 ← BI	0.796		0.855	0.73	1596.189 (p = 0.000)
BI2 ← BI	0.858	29.41 ***
BI3 ← BI	0.792	27.915 ***
BA4 ← BAs	0.795		0.892	0.842	2708.109 (p = 0.000)
BA3 ← BAs	0.825	30.634 ***
BA2 ← BAs	0.799	29.793 ***
BA1 ← BAs	0.865	32.099 ***

Note: *** p < 0.001, two-tailed.

) [49].

5.2. Model Fitting and Fit Testing

Based on the model hypothesis, measurement index design, and the results of reliability and validity tests, a structural equation model was constructed with SNs, BAs, PEC, BI, and BR as latent variables, as illustrated below (Figure 2). The cognitive–behavioral model of garbage sorting was assessed using AMOS 26.0 software. Given reasonable covariance relationships identified among the variables’ variances, six groups of covariance relationships—e11 vs. e2, e8 vs. e14, e9 vs. e14, e9 vs. e10, e10 vs. e15, and e11 vs. e15—were incorporated based on preliminary results. This approach aimed to reduce the model’s chi-square value without violating theoretical assumptions, thereby enhancing the model’s goodness of fit [44].

Fitness indices evaluate how well a hypothetical path analysis model fits the sample data. The absolute fit index and the incremental fit index are commonly used to measure the overall goodness-of-fit of the structural equation model. As indicated in the table below, the CFI value of this model is 0.982, surpassing the threshold of 0.900 [44]. Additionally, the statistical values for the remaining goodness-of-fit indices meet or approach their respective thresholds, demonstrating that the structural equation model constructed in this study exhibits a strong fit, and thus the model’s overall fitness successfully passes the test (

Table 4. Indicators of model fit.

Statistical Test		Criteria	Measured Value
Absolute goodness-of-fit indicators	CMIN/DF	The ratio of the chi-square value to the degrees of freedom < 3 as a criterion that the model has a good fit	2.854
	RMSEA	Closer to 0 indicates better model fit, usually using RMSEA < 0.08	0.04
Value-added goodness-of-fit indicators	NFI	Closer to 1 indicates a better model fit, usually using NFI > 0.90	0.972
	RFI	Closer to 1 indicates a better model fit, usually using an RFI > 0.90	0.963
	IFI	Closer to 1 indicates a better model fit, usually using an IFI > 0.90	0.982
	TLI	Closer to 1 indicates a better model fit, usually using a TLI > 0.90	0.976
	CFI	Closer to 1 indicates a better model fit, usually using a CFI > 0.90	0.982

) [44].

5.3. Analysis

The structural equation modeling (SEM) results with bootstrap testing confirmed the proposed hypotheses, demonstrating significant mediation effects.

In the structural equation model (

Table 5. Estimation results of structural equation modeling.

Trails	Standardized Path Factor	Standard Error	Critical Ratio
BI ← SNs	0.2 ***	0.038	5.206
BI ← BAs	0.179 ***	0.034	5.106
BI ← PEC	0.25 ***	0.039	6.906
BR ← SNs	0.188 ***	0.036	4.94
BR ← BAs	0.218 ***	0.033	6.277
BR ← PEC	0.186 ***	0.037	5.118
BR ← BI	0.189 ***	0.034	5.333
SNs ↔ BAs	0.449 ***	0.044	11.76
SNs ↔ PEC	0.402 ***	0.041	10.471
Bas ↔ PEC	0.270 ***	0.038	7.585

Note: *** p < 0.001, 2-tailed.

), the path coefficient of subjective norms (SNs) on individual willingness (BI) is 0.2, significant at the 0.001 level. This indicates that the subjective norms of rural residents significantly influence individual behavioral willingness. Specifically, a one-unit increase in subjective norm scores corresponds to a 20% increase in behavioral intention. Hypothesis H1 is supported. The path coefficient of behavioral attitude (BAs) on personal intention (BI) is 0.179, significant at the 0.001 level, demonstrating that behavioral attitude crucially affects personal intention. A one-unit increase in behavioral attitude corresponds to a 17.9% increase in behavioral intention. Hypothesis H3 is supported. The path coefficient of perceived behavioral control (PEC) on personal intention (BI) is 0.25, significant at the 0.001 level. This shows that perceived behavioral control significantly influences personal intention. Perceived behavioral control exerted the strongest positive effect on behavioral intention among the three cognitive factors, with a one-unit increase in PBC corresponding to a 25% increase in sorting intention. Hypothesis H5 is supported.

Meanwhile, the path coefficient of subjective norms (SNs) on behavioral response (BR) is 0.188, significant at the 0.001 level, demonstrating that subjective norms significantly influence behavioral responses. A one-unit increase in subjective norms corresponds to an 18.8% increase in behavioral response for waste sorting. Hypothesis H2 is supported. The path coefficient of behavioral attitude (BAs) on behavioral response (BR) is 0.218, significant at the 0.001 level, indicating that behavioral attitude significantly influences behavioral responses. For every one-unit increase in villagers’ behavioral attitude, their behavioral response to waste sorting increases by 21.8%. Hypothesis H4 is supported. The path coefficient of perceived behavioral control (PEC) on behavioral response (BR) is 0.186, significant at the 0.001 level, showing that perceived behavioral control strongly influences behavioral responses. A one-unit increase in perceived behavioral control corresponds to an 18.6% increase in behavioral response for waste sorting. Hypothesis H6 is supported. Meanwhile, the path coefficient of personal willingness (BI) on behavioral response (BR) is 0.189, significant at the 0.001 level, suggesting that personal willingness critically influences behavioral responses. For every one-unit increase in villagers’ individual intention, their behavioral response to waste sorting increases by 18.9%. Hypothesis H7 is supported. Moreover, the results indicate that behavioral intention mediates the relationship between cognitive factors and behavioral response in rural waste sorting. Hypothesis H8 is supported. Specifically, behavioral intention accounts for 3.8%, 3.4%, and 4.7% of the indirect effects of subjective norm, behavioral attitude, and perceived behavioral control on behavioral response, respectively (

Table 6. Analysis of intermediation effects.

Effect (Scientific Phenomenon)	SNs → BI → BR	Bas → BI → BR	PBC → BI → BR	BI → BR
Direct effect	0.188 (p = 0.000)	0.218 (p = 0.000)	0.186 (p = 0.000)	0.189 (p = 0.000)
Indirect effect	0.038 (p = 0.000)	0.034 (p = 0.000)	0.047 (p = 0.000)
Aggregate effect	0.226 (p = 0.000)	0.252 (p = 0.000)	0.233 (p = 0.000)	0.189 (p = 0.000)

).

5.4. Discussion

Additionally, the empirical analysis using structural equation modeling (SEM) revealed a two-by-two correlation among individual subjective norms, behavioral attitudes, and perceived behavioral control. This finding does not render our overall results inconclusive. Rather, it underscores the complex interplay between these three constructs and indicates that further pathways or interactions may exist beyond the direct effects and mediation paths tested in our model. For instance, subjective norms and behavioral attitudes might reinforce one another, or perceived behavioral control could partially moderate the relationship between norms and attitudes. Further research is required to explore this issue.

These possibilities suggest that rural residents’ responses to waste classification may be shaped by multi-layered influences. When villagers perceive strong social pressure (subjective norms) but lack the necessary resources or confidence to carry out waste sorting (low perceived behavioral control), their positive attitudes may not translate into consistent actions. Conversely, high perceived behavioral control may strengthen the effect of positive attitudes, leading to greater consistency between willingness and behavior.

To elucidate these alternate pathways, future research could employ extended theoretical frameworks (e.g., integrating elements from the Value–Belief–Norm (VBN) theory or incorporating additional moderators, such as trust in local governance). Longitudinal or multi-wave panel data could capture how these variables evolve over time, allowing researchers to pinpoint dynamic interactions and directional influences among subjective norms, attitudes, and perceived control. Such studies would complement and refine the insights gained from our present cross-sectional analysis, ultimately deepening the understanding of how rural households adopt and sustain waste classification behavior. Moreover, future research should investigate the intention–behavior gap and the moderating effects of external factors on waste-sorting intentions and behaviors to advance the literature on rural waste-sorting governance.

Overall, while our current findings highlight the importance of subjective norms, behavioral attitudes, and perceived control, the two-by-two correlations encourage us to view these constructs as interdependent rather than wholly discrete factors. Recognizing this interdependence not only broadens the theoretical scope of the analysis but also paves the way for more targeted policy interventions, such as combined education and infrastructure improvements that strengthen perceived control while simultaneously reinforcing positive attitudes and supportive social norms.

6. Conclusions and Recommendations

This study employs the Theory of Planned Behavior (TPB) and structural equation modeling (SEM) to investigate the full decision-making process of rural residents’ household waste classification in Jiangsu Province, China. By examining the entire chain from cognition to behavior in a large-scale sample of 2420 rural households, the research provides a more comprehensive understanding of waste-sorting decision making than is typically captured in urban-focused studies. The empirical results demonstrate that subjective norms, behavioral attitudes, and perceived behavioral control each exert a significant and positive influence on both behavioral intentions (willingness) and actual classification behaviors. Furthermore, the mediating effect of behavioral intentions underscores the central TPB proposition that intention is the critical link between cognition and action. Overall, our model’s strong fit indices and the alignment of theoretical assumptions with real-world data reinforce TPB’s validity in explaining the complex dynamics of rural waste classification.

Based on these findings, this study offers three recommendations. First, it is essential to refine and extend policy frameworks. Existing regulations and classification guidelines often overlook the particularities of rural environments, which can dampen residents’ motivation. To encourage more consistent sorting, policymakers must promulgate legal provisions specifically adapted to rural needs, establish clear and unambiguous waste-sorting standards, enhance training programs for farmers on proper waste classification, and support locally adjustable disposal systems. Second, strengthening villagers’ agency through integrated incentive and supervision mechanisms is crucial for mobilizing villagers’ subjective initiative. Structural equation modeling reveals that behavioral attitudes significantly influence waste-sorting behaviors, highlighting the importance of motivating villagers’ participation. Firstly, grassroots organizations should demonstrate leadership by having village officials and party members model proper waste-sorting practices. Secondly, positive incentive mechanisms should be established by linking waste-sorting performance with access to collective benefits and honorary recognitions. Finally, self-governance supervision systems should be enhanced through regular inspections and public reporting of results to foster sustained behavioral change. Third, social norms should be established through policy guidance and peer influence. Effective waste sorting requires coordinated policy interventions and social norm construction. Fourth, targeted policy campaigns should employ culturally appropriate methods to enhance environmental awareness. Fifth, peer influence mechanisms should be activated through volunteer teams and mutual-assistance groups to create demonstration effects. Concurrently, village regulations should incorporate appropriate sanctions for non-compliance, fostering community-wide participation in waste-sorting practices.

Overall, this study validates the application of TPB within rural China and underscores the importance of embedding theoretical insights into well-designed local policies. Future research may consider integrating extended concepts—such as moral norms or technology acceptance perspectives—and employing longitudinal methods to track how waste-classification behaviors evolve over time. Through continued investigation and targeted policy innovation, rural communities can enhance the efficacy of domestic waste management and contribute to broader sustainability goals.