Policy Tools, Policy Perception, and Compliance with Urban Waste Sorting Policies: Evidence from 34 Cities in China

Abstract

Promoting municipal solid waste (MSW) sorting is critical to advancing sustainable and low-carbon urban development. While existing research often focuses separately on external policy tools or internal behavioral drivers, limited attention has been given to their joint effects within an integrated framework. This study addresses this gap by analyzing micro-survey data from 1983 residents across 34 prefecture-level and above cities in China, using a bivariate probit model to examine how policy tools and policy perception—both independently and interactively—shape residents’ active and passive compliance with MSW sorting policies. The findings reveal five key insights. First, the adoption and spatial distribution of policy tools are uneven: environment-type tools dominate, supply-type tools are moderately deployed, and demand-type tools are underutilized. Second, both policy tools and policy perception significantly promote compliance behaviors, with policy cognition exerting the strongest effect. Third, differential effects are observed—policy cognition primarily drives active compliance, whereas policy acceptance more strongly predicts passive compliance. Fourth, synergistic effects emerge when supply-type tools are combined with environment-type or demand-type tools. Finally, policy perception not only directly enhances compliance but also moderates the effectiveness of policy tools, with notable heterogeneity among residents with higher cognitive or emotional alignment. These findings contribute to a deeper understanding of compliance mechanisms and offer practical implications for designing perception-sensitive and regionally adaptive MSW governance strategies.

1. Introduction

Municipal solid waste (MSW) sorting is essential for sustainable urban development, delivering benefits such as reduced resource consumption, improved environmental quality, and support for circular economy goals. Urbanization is accelerating this challenge—55% of the global population now lives in cities, and this figure is projected to rise to 66% by 2050, driving global MSW generation from approximately 2.1 billion tonnes in 2023 to a projected 3.8 billion tonnes by mid-century [1]. Despite growing policy attention, less than 20% of global waste is recycled annually, leaving over 1.7 billion tonnes unmanaged, which imposes severe environmental pressures and strains municipal budgets worldwide [2]. In response, many governments have adopted source-based MSW sorting policies to reduce waste at the point of generation and to align waste management systems with broader climate mitigation and low-carbon development goals [3,4].

Countries such as Sweden, EU-member states, and China have launched ambitious municipal solid waste (MSW) sorting initiatives, albeit through differing approaches and institutional capacities [5]. In Sweden and other EU-member states, well-established extended producer responsibility (EPR) systems and long-standing public awareness campaigns have resulted in MSW recycling rates exceeding 45%, with some Nordic countries achieving over 99% diversion from the landfill through advanced incineration and material recovery processes [6]. In contrast, China’s top–down policy-driven approach has prioritized the large-scale rollout of mandatory sorting policies, which by 2021 had reached 297 prefecture-level cities—covering approximately 150 million urban households, or around 45% of the urban population [7]. However, public participation remains uneven in China, with studies reporting what has been termed “government enthusiasm but public apathy” [8]. Surveys reveal a disconnect between formal policy mandates and actual household behavior, highlighting the challenges of fostering behavioral compliance in the absence of bottom–up engagement and systemic incentives.

Most existing studies on MSW sorting behavior draw on behavioral models—such as the Attitude–Behavior–Context (ABC) model [9], Theory of Planned Behavior [10], and Norm Activation Theory [11]—to explore determinants including environmental concern, incentive schemes, social capital [12,13,14], and infrastructure quality [15]. Among these, the traditional Knowledge–Attitude–Practice (KAP) model also provides a foundational lens by linking information awareness, attitudinal disposition, and behavioral execution [16]. However, recent studies have challenged its linear and rationalist assumptions, arguing that knowledge does not automatically lead to action, especially in complex institutional settings or emotionally driven contexts [17,18]. Yet, few have examined how top–down policy tools and bottom–up public perceptions interact to shape compliance [19,20]. In particular, the ways in which residents’ policy perception and specific policy tools jointly influence active versus passive compliance behaviors remain under-researched in the Chinese context.

To address this gap, our study analyzes data from 1983 respondents across 34 prefecture-level cities in China, constructing a theoretical framework that integrates policy tools and residents’ policy perception. We apply a bivariate probit model to test their independent, interactive, and moderated impacts on both active and passive compliance. Our results not only uncover a nuanced compliance mechanism but also offer evidence-based recommendations for tailored, perception-aware policy interventions to support sustainable waste governance [21,22].

The remainder of the paper is organized as follows: Section 2 reviews the theoretical background and hypotheses; Section 3 details materials and methods; Section 4 presents empirical results; Section 5 discusses the findings; Section 6 concludes; and Section 7 proposes recommendations based on our analysis.

2. Theoretical Background and Hypotheses

Policy compliance refers to the extent to which individuals adhere to the rules and expectations set by public policy, often influenced by perceptions of legitimacy, enforcement, and perceived benefit streams [23,24]. In the context of municipal solid waste (MSW) sorting, the effectiveness of governance largely hinges on the degree of front-end compliance among residents. However, in the absence of external intervention, many urban residents tend to default to the more convenient practice of mixed waste disposal, consistent with the principle of minimizing personal cost [25]. From a rational choice perspective, individuals evaluate the potential benefits and penalties associated with compliance and opt for actions that maximize their utility [26]. This perspective underpins the “rational economic man” model, which views compliance as a function of contextual factors and instrumental incentives embedded in policy tools [27].

Yet, real-world behavior is also shaped by bounded rationality and heuristic decision-making. Residents do not always act as purely rational agents; their compliance decisions are influenced by perceptions, emotions, and cognitive shortcuts [28,29]. Accordingly, the “behavioral person” model highlights the role of subjective judgments, including policy-related attitudes and emotional responses [30,31]. From this standpoint, compliance is not only a cost–benefit calculation but also a reflection of individuals’ policy cognition and acceptance. Synthesizing these views, this study develops a dual-perspective framework for MSW policy compliance, integrating the macro-level influence of policy tools with the micro-level dynamics of policy perception (see Figure 1).

2.1. The Impact of Policy Tools on Urban Residents’ Compliance with MSW Sorting Policies

Urban residents’ MSW sorting is an external behavior that not only impacts the individual or household but also significantly affects social environmental protection and resource utilization. However, its impact on third parties is not reflected by market price mechanisms. To correct this “market failure”, Arthur Pigou’s externality theory advocates using policy tools to internalize externalities, enabling social costs and benefits to be considered in individual decisions [32,33]. Specifically, with the use of different types of policy tools (such as taxes, subsidies, or regulations), the external effects of MSW sorting can be included in individual decision-making, effectively promoting MSW sorting among residents [34].

Based on Rothwell and Zegveld’s classification [35], policy instruments can be divided into supply-type, demand-type, and environment-type categories [36]. Supply-type tools include infrastructure provision, technical support, financial subsidies, and related resources aimed at lowering barriers to residents’ compliance. Demand-type tools encompass incentives, penalties, educational campaigns, and participatory initiatives designed to motivate active resident engagement. Environment-type tools involve regulatory frameworks, clear policy targets, supervision mechanisms, and standards aimed at creating a stable and conducive governance environment. Prior empirical evidence confirms the positive role of these policy tools in influencing residents’ MSW sorting compliance [37,38].

Consequently, the following hypotheses are proposed:

Hypothesis 1 (H1).  

Policy tools exert a direct positive impact on urban residents’ compliance with MSW sorting policies.

Hypothesis 1a (H1a).  

Supply-type tools exert a direct positive impact on compliance.

Hypothesis 1b (H1b).  

Environment-type tools exert a direct positive impact on compliance.

Hypothesis 1c (H1c).  

Demand-type tools exert a direct positive impact on compliance.

2.2. The Impact of Policy Perception on Urban Residents’ Compliance with MSW Sorting Policies

Policy perception refers to how individuals interpret, assess, and accept public policies, shaping their behavioral responses accordingly [39]. A well-informed and positively engaged public is more likely to exhibit compliance, whereas misinformation or distrust can hinder alignment with policy goals [40]. Policy perception thus plays a crucial role in bridging the gap between policy intent and behavioral implementation. This study conceptualizes policy perception through two dimensions:

Policy cognition, which refers to residents’ knowledge and understanding of the policy’s content, objectives, implementation process, and expected outcomes. The Theory of Planned Behavior (TPB) suggests that greater policy cognition leads to stronger behavioral intentions and higher compliance likelihood. Accurate cognition enables individuals to form reasoned judgments, enhancing the likelihood of compliance with public policies [41].

Policy acceptance, which reflects the degree to which residents emotionally identify with and normatively support MSW sorting policies, enhances voluntary engagement, internalizes sustainable behaviors, and contributes to collective efficacy in environmental governance [42]. When individuals perceive a policy as fair, effective, and consistent with shared social values, they are more inclined to comply [43].

Accordingly, we propose the following:

Hypothesis 2 (H2).  

Policy perception has a direct positive effect on policy compliance.

2.3. The Interaction Effects Between Policy Tools and Policy Perception

This section explores the interactive effects between policy tools and policy perception in shaping urban residents’ compliance with MSW sorting policies. Specifically, two forms of moderation are examined:

First, policy perception may moderate the effectiveness of policy tools. When residents perceive sorting policies as legitimate, reasonable, and beneficial, the impact of supply-type tools, demand-type tools, or environment-type tools is likely to be amplified. High levels of perceived legitimacy and policy approval enhance the credibility and acceptability of government interventions, thereby increasing voluntary compliance [44].

Second, policy tools may also influence policy perception. Well-designed and consistently implemented tools, such as clear regulations, accessible infrastructure, and responsive public services, can enhance residents’ understanding of policy content, reduce ambiguity, and foster trust in public institutions [45]. These mechanisms, in turn, reinforce policy cognition and acceptance, which are essential components of compliance behavior.

By improving both the instrumental and cognitive dimensions of resident engagement, policy tools and perception may form a mutually reinforcing system that facilitates compliance with public environmental mandates.

Accordingly, we propose the following hypotheses:

Hypothesis 3 (H3).  

Policy perception positively moderates the relationship between policy tools and policy compliance.

Hypothesis 4 (H4).  

Policy tools positively moderate the relationship between policy perception and policy compliance.

3. Materials and Methods

3.1. Sample Source

The microdata of residents used in this study were collected from a survey conducted by the research team between July 2022 and May 2024. The survey covered 34 prefecture-level cities across 12 provinces in eastern (Jiangsu, Zhejiang, Guangdong), central (Hubei, Henan, Anhui), western (Sichuan, Shaanxi, Yunnan), and northeastern (Heilongjiang, Jilin, Liaoning) China. These regions differ significantly in geographical location, economic development level, and cultural background, and their consideration supports a comprehensive understanding of urban residents’ compliance with MSW sorting policies in China.

The survey used a combination of stratified sampling and random sampling. A total of 2500 structured questionnaires were distributed during field investigations, based on the quotas for residents’ gender, monthly income, and education level. A total of 2389 questionnaires were collected, for a recovery rate of 95.56%. After the exclusion of invalid samples (e.g., blank questionnaires, those with repeated options, and screening questions), 1983 valid questionnaires were obtained, for an effective rate of 79.32%. Policy data were sourced from policy resources such as the “National Laws and Regulations Database”, “PKU Law”, and government websites. MSW sorting policies from the 34 surveyed cities were selected as observation samples. Policies related to MSW sorting issued between March 2017 (when the national “Implementation Plan for MSW Sorting System” was introduced) and May 2024 were screened, resulting in 70 valid policy documents for analysis.

3.2. Policy Document Collection

To complement the survey data and empirically assess the role of policy tools, we systematically collected municipal solid waste (MSW) sorting policy documents from official and authoritative sources, including the National Laws and Regulations Database, PKU Law Database, and websites of municipal and provincial governments. The selection covered the same 34 prefecture-level cities included in the resident survey, ensuring consistency between policy context and behavioral data.

The time frame for policy collection spanned from March 2017, marking the release of the national Implementation Plan for MSW Sorting Systems, to May 2024, thereby capturing the evolution of local governance practices over a seven-year period. After screening out duplicates, internal working documents, and policies not directly related to household waste sorting, a total of 70 valid policy documents were retained for analysis.

For analytical consistency, these documents were coded using a threefold typology of policy tools: supply-type, demand-type, and environment-type (see Section 2.1), following established classifications from Rothwell and Zegveld and Howlett and Ramesh [35,36]. This coding provided the empirical basis for evaluating the policy mix and examining its interaction with resident perception in shaping compliance behavior.

3.3. Variable Selection

3.3.1. Policy Compliance

In line with the existing literature on compliance behavior [46], this study conceptualizes policy compliance as a binary behavioral response to municipal solid waste (MSW) sorting policies [47]. Two dimensions are distinguished: Active compliance, reflecting voluntary adherence to policy rules, was measured by the question: “Do you actively comply with MSW sorting policies?” Responses were coded as 1 for “yes” and 0 for “no”. Passive compliance, indicating habitual compliance without external pressure, was assessed through the counterfactual question: “If there were no policy management and supervision, would you still sort waste?” A “yes” response was coded as 1, and “no” as 0. This dual measure captures both motivation and behavioral intention, aligning with the theoretical emphasis on rational action and bounded rationality discussed in Section 2.

3.3.2. Policy Tools

The classification of policy tools follows the typology established by Rothwell and Zegveld and Howlett and Ramesh, dividing instruments into supply-type, demand-type, and environment-type categories. Content analysis of policy documents from the surveyed 34 cities was conducted, and the intensity of each policy type was quantified by counting the number of policy items (nodes) per category. These counts served as continuous variables, enabling empirical analysis of their direct and moderating influences on residents’ compliance behaviors.

3.3.3. Policy Perception

Policy perception was operationalized through two constructs: Policy cognition is defined as residents’ knowledge and understanding of MSW sorting policies, including objectives, content, and expected outcomes. Following Ajzen’s Theory of Planned Behavior (1991) and based on the scale developed by Wang et al. [48], responses were captured using a 5-point Likert scale and reduced to a composite index through principal component analysis (PCA). Policy acceptance is defined as the degree of emotional and normative support for MSW sorting policies. This variable was measured using a 4-item scale adapted from Burak, M et al. [49], also rated on a 5-point Likert scale and processed via PCA.

To account for potential proximity bias, a screening question was included: “Have you participated in the formulation of China’s or your city’s MSW sorting policies?” (Yes = 1, No = 0).

3.3.4. Control Variables

To reduce potential omitted variable bias and account for individual heterogeneity in policy compliance behavior, this study incorporates a series of control variables commonly used in environmental behavior research [50]. These variables capture residents’ demographic profiles, socioeconomic status, and levels of governance participation, all of which may influence their capacity or motivation to comply with MSW sorting policies. The control variables include age, gender, health status, employment, housing ownership, income level, education, and participation in governance or publicity activities. Each variable reflects potential factors influencing residents’ motivation or capacity to comply with MSW sorting policies. Detailed definitions and coding of control variables are provided in

Table 1. Variable assignment and descriptive statistics.

	Variables	Description	Assignment	Mean Value	Standard Deviation
Dependent variable	Active policy compliance	Are you proactively complying with MSW sorting policies?	Yes = 1, No = 0	0.57	0.50
	Passive policy compliance	If there were no relevant policy management and supervision, would you not comply with MSW sorting?	No = 1, Yes = 0	0.62	0.49
Policy Tools	Environment-type tools	Policy text results in terms of the number of corresponding nodes		41.36	25.49
	Supply-type tools			22.62	12.90
	Demand-type tools			16.81	7.66
Policy Perception	Policy cognition	Have you heard of the MSW sorting policy?	1 = Strongly Disagree; 2 = Disagree; 3 = Neutral; 4 = Agree; 5 = Strongly Agree	4.02	1.07
		Are you familiar with the specific policies and implementation methods of MSW sorting?		4.01	1.10
		Do you believe that local governments should ensure the advancement of residential solid waste sorting and the construction of treatment facilities?		4.13	1.11
		Do you believe that MSW sorting policies can achieve waste reduction, resource utilization, and harmless treatment?		4.01	1.10
	Policy acceptance	Do you accept the MSW sorting policies and policy framework in your city?		3.70	1.16
		Do you support the MSW sorting policy and policy system in your city?		3.77	1.22
		Are you satisfied with the performance of the MSW sorting policy and system in your city?		3.89	1.26
		Are you optimistic about the effectiveness of MSW sorting policies in your city?		3.76	1.25
Control Variables	Age	1 = 0–18 years old; 2 = 19–40 years old; 3 = 41–65 years old; 4 = above 65 years old		2.50	0.69
	Gender	1 = Male; 0 = Female		0.35	0.48
	Health status	1 = Healthy; 0 = Unhealthy		0.77	0.42
	Employment status	1 = Employed; 0 = Unemployed		0.90	0.31
	Housing status	1 = Self-owned housing; 0 = Non-self-owned housing		0.71	0.45
	Income level	Annual disposable income (unit: RMB)		1.91	0.74
	Education level	1 = Primary School; 2 = Middle School; 3 = High School/Vocational School; 4 = Associate Degree; 5 = Bachelor’s Degree; 6 = Master’s Degree; 7 = Doctoral Degree		4.18	1.25
	Participation in community governance	1 = Participated; 0 = Not participated		0.81	0.39
	Involvement in publicity and training	1 = Participated; 0 = Not participated		0.82	0.39
Instrumental Variable	The accuracy of MSW sorting	The ability to accurately sort household waste	Yes = 1, No = 0	0.77	0.42
Screening criteria	Psychological distance	Have you participated in the formulation of waste sorting policies in China or your city?	Yes = 1, No = 0	0.00	0.00

.

3.3.5. Instrumental Variable

To address potential endogeneity between policy perception and policy compliance, this study introduces an instrumental variable (IV): the accuracy of MSW sorting, which measures residents’ ability to correctly classify household waste types. This variable is coded as a binary indicator (1 = accurate sorting; 0 = inaccurate). The selection of this variable as an instrument is grounded in two key considerations. First, sorting accuracy is highly correlated with residents’ cognitive understanding of MSW sorting policies, thereby satisfying the relevance condition for a valid instrument. Individuals who can accurately sort waste are more likely to be familiar with policy content, implementation processes, and expected outcomes—core components of policy perception [51]. Second, sorting accuracy is not directly driven by motivational or normative factors associated with compliance behavior, especially when hypothetical counterfactuals (e.g., absence of monitoring) are considered [52]. Hence, it fulfills the exclusion restriction required for instrumental variable estimation. The accuracy of MSW sorting (mean = 0.77; SD = 0.42), presented in Table 1, serves as an instrumental variable for estimating the causal effect of policy perception on compliance behavior.

3.4. Model Construction

A bivariate probit (Biprobit) model was employed to simultaneously estimate the determinants of active and passive compliance with municipal solid waste (MSW) sorting policies. Compared to separate univariate probit models, this approach accounts for potential error term correlation between the two compliance behaviors, thereby addressing unobserved heterogeneity and mitigating endogeneity concerns, resulting in more efficient and consistent parameter estimates. The decision variables for the active and passive policy compliance of residents are $Y_{1i}$ and $Y_{2i}$, respectively. For each resident i, the bivariate probit model is constructed as follows:

$$\left\{\begin{array}{c} Y_{1i}^{\mathrm{*}}=\alpha X_i+{\epsilon }_i \\ Y_{2i}^{\mathrm{*}}=\beta X_i+{\mu }_i\end{array}\right.$$

(1)

where $Y_{1\mathrm{i}}^{\mathrm{*}}$ and $Y_{2\mathrm{i}}^{\mathrm{*}}$ are latent variables for active and passive policy compliance, respectively; $X_i$ represents the factors influencing policy compliance, including policy tools and policy perception; $\mathsf{\alpha }$ and $\mathsf{\beta }$ are coefficient vectors; and ${\epsilon }_i$ and ${\mu }_i$ are random disturbances, following a bivariate joint normal distribution with a covariance of $\rho$, that is, $\left(\begin{array}{c}{\epsilon }_i\\ {\mu }_i\end{array}\right)~N\left\{\left(\begin{array}{c}0\\ 0\end{array}\right),\left[\begin{array}{c}1\\ \rho \end{array}\right.\left.\begin{array}{c}\rho \\ 1\end{array}\right]\right\}$.

The observed variables $Y_{1\mathrm{i}}$ and $Y_{2\mathrm{i}}$ follow specific decision rules: if $Y_{1\mathrm{i}}^{\mathrm{*}}$ > 0, then $Y_{1\mathrm{i}}$ = 1 indicates active compliance; otherwise, $Y_{1\mathrm{i}}$ = 0. If $Y_{2\mathrm{i}}^{\mathrm{*}}$ > 0, then $Y_{2\mathrm{i}}$ = 1 indicates passive compliance; otherwise, $Y_{2\mathrm{i}}$ = 0.

For the sample where both active and passive compliance occur, the corresponding ${\rho }_{11}$ is:

$$\begin{gathered} {\rho }_{11}=P(Y_{1i}=1,Y_{2i}=1)=P(Y_{1i}^{\mathrm{*}}>0,Y_{2i}^{\mathrm{*}}>0)=P({\epsilon }_i>-\alpha X_i,{\mu }_i \\ >-\beta X_i)\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em} \\ =P({\epsilon }_i<\alpha X_i,{\mu }_i<\beta X_i)={\int }_{-\mathrm{\infty }}^{\alpha X_i}{\int }_{-\mathrm{\infty }}^{\beta X_i}\phi (z_1,z_2,\rho )d{z}_{1}d{z}_2\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em} \\ =\varphi (\alpha X_i,\beta X_i,\rho )\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em} \end{gathered}$$

(2)

where $P(·)$ denotes the probability; $\phi (·)$ denotes the probability density function; $\varphi (·)$ denotes the cumulative distribution function; and $z_1$, $z_2$ is the integration variable. The same logic applies to ${\rho }_{10}$, ${\rho }_{01}$, and ${\rho }_{00}$. Since the bivariate probit model is nonlinear, it cannot be transformed into a linear model. Therefore, the maximum likelihood estimation (MLE) method is used to jointly estimate ${\rho }_{11}$, ${\rho }_{10}$, ${\rho }_{01}$, and ${\rho }_{00}$, and the log-likelihood function is obtained by summing the logarithms of these probabilities:

$$\begin{gathered} \mathit{ln}L\left(\alpha ,\beta ,\rho \right)={\sum }_{i=1}^n\left\{Y_{1i}{Y}_{2i}\mathit{ln}F\left(\alpha X_i,\beta X_i;\rho \right)+Y_{1i}\left(1-Y_{2i}\right)\mathit{ln}\left[\varphi \left(\alpha X_i\right)- \\ F\left(\alpha X_i,\beta X_i,\rho \right)\right]+\left(1-Y_{1i}\right)\mathit{ln}\varphi \left(-\alpha X_i\right)\right\} \end{gathered}$$

(3)

4. Results

4.1. Selection and Spatial Distribution of Policy Tools for MSW Sorting in Chinese Cities

Through content analysis and categorization of municipal solid waste (MSW) sorting policy documents from 34 prefecture-level cities across China, we analyzed the distribution of three policy tool categories: environment-type, supply-type, and demand-type (see Figure 2). Overall, the distribution is significantly uneven. Environment-type tools comprise the largest proportion (51.35%, n = 1405), followed by supply-type tools (27.96%, n = 765), with demand-type tools being the least utilized (20.69%, n = 566). This distribution pattern indicates a governance strategy characterized by strong reliance on environmental-side tools, moderate use of supply-side tools, and limited application of demand-side tools.

Specifically, the findings include the following: (1) Strong reliance on environment-type tools: Among these tools, legal regulations constitute the largest proportion (43.84%), followed by responsibility division (16.94%), supervision and inspection (12.10%), evaluation and assessment (8.68%), target planning (11.46%), and standard norms (6.98%). This structure reveals the local governments’ preference for mandatory and supervisory policy tools, reflecting a governance approach focused on compliance through authoritative mechanisms [53]. However, the relatively lower proportions of evaluation and assessment, along with standard norms, suggest insufficient development of comprehensive and detailed management frameworks. (2) Moderate reliance on supply-type tools: Within supply-type instruments, local governments exhibit a clear preference for visible and tangible interventions, with public services (35.42%) and infrastructure development (28.50%) dominating the portfolio, followed by financial support (11.50%), personnel allocation (10.98%), technical guidance (9.80%), and minimal land input (3.79%). Here, land input refers to the allocation of urban space for MSW sorting infrastructure—such as drop-off points, recycling stations, or transfer depots. The relatively low use of spatial interventions suggests that land-based planning remains underutilized, potentially limiting accessibility, scalability, and the long-term institutionalization of sorting systems in dense urban areas [54]. (3) Weak reliance on demand-type tools: Publicity and training represent the largest share (45.23%) within demand-side instruments, followed by demonstration projects (13.96%), participatory management (13.96%), public–private partnerships (12.37%), incentives and penalties (11.84%), and domain-specific projects (2.65%). This structure highlights insufficient market-driven incentives and limited public engagement, reflecting the early development stage of multi-stakeholder governance mechanisms [55]. The comparatively low utilization of public–private partnerships and domain-targeted strategies indicates an early stage of institutional development in demand-side MSW governance frameworks [56,57].

From a regional perspective, significant variations in the selection and application of policy tools were observed across the four regions (eastern, central, western, and northeastern) of China (see Figure 3).

The western region prominently employs environment-type tools (53%) due to its strategic ecological importance within China’s national ecological security strategy (e.g., the “Two Screens and Three Belts” initiative). Local governments in cities such as Chengdu and Xi’an have implemented stringent regulatory measures, including integrating compliance records into social credit systems, to ensure effective enforcement and reduce environmental degradation associated with rapid urbanization.

In the northeastern region, environment-type tools also dominate (48%), primarily due to ongoing economic structural transformations. With limited fiscal capacity, cities such as Fushun tend to prioritize regulatory instruments that are cost-effective and administratively feasible. This pattern reflects not only local fiscal constraints but also a broader governance logic observed in other post-industrial or transitioning cities—where low-capex, high-enforcement tools are preferred over resource-intensive infrastructure investments [58,59]. These findings suggest that regional tool preferences are shaped by both economic conditions and administrative capacity, reinforcing the importance of aligning tool deployment with local development trajectories [60].

Conversely, the eastern and central regions, benefiting from relatively developed economies and stronger institutional capacities, exhibit a balanced reliance on supply-type tools (approximately 30%) and a more substantial application of infrastructure and technological innovations. For instance, Hangzhou has adopted advanced MSW sorting robots through public–private collaborations, reflecting their ability to afford technological investments and provide higher-quality public services. Such an approach illustrates the adaptability and efficacy of supply-side tools in economically prosperous regions.

These spatial variations underscore the importance of region-specific governance strategies that consider local economic conditions, environmental priorities, and institutional capacities, thereby enhancing the effectiveness and sustainability of MSW sorting policies across diverse urban contexts in China.

4.2. Analysis of the Impact of Policy Tools and Policy Perception on Urban Residents’ Compliance with MSW Sorting Policies

To empirically examine the influence of policy tools and policy perception on urban residents’ compliance with MSW sorting policies, a bivariate probit model was utilized.

Table 2. Effects of policy tools and policy perception on residents’ compliance behavior with municipal solid waste sorting policies.

Variables	Biprobit		Marginal Effect
	Active Policy Compliance	Passive Policy Compliance	Active Policy Compliance	Passive Policy Compliance	Simultaneous Policy Compliance
Environment-type	0.004 (0.003)	0.013 *** (0.004)	0.001 (0.001)	0.003 *** (0.001)	0.002 *** (0.001)
Supply-type	0.017 *** (0.006)	0.022 *** (0.007)	0.004 *** (0.002)	0.005 *** (0.002)	0.006 *** (0.001)
Demand-type	0.015 ** (0.007)	−0.004 (0.008)	0.004 ** (0.002)	−0.001 (0.002)	0.002 (0.002)
Policy cognition	0.549 *** (0.073)	0.197 *** (0.056)	0.144 *** (0.018)	0.045 *** (0.013)	0.116 *** (0.013)
Policy acceptance	0.170 *** (0.045)	0.361 *** (0.045)	0.044 *** (0.012)	0.082 *** (0.010)	0.073 *** (0.009)
Age	−0.140 *** (0.051)	−0.256 *** (0.053)	−0.037 *** (0.013)	−0.059 *** (0.012)	−0.055 *** (0.011)
Gender	−0.037 (0.072)	−0.036 (0.077)	−0.010 (0.019)	−0.008 (0.018)	−0.011 (0.016)
Health status	0.377 *** (0.091)	0.624 *** (0.096)	0.098 *** (0.024)	0.141 *** (0.021)	0.141 *** (0.020)
Employment status	−0.081 (0.126)	0.177 (0.122)	−0.021 (0.033)	0.040 * (0.028)	0.009 (0.027)
Housing status	0.104 (0.079)	0.081 (0.085)	0.027 (0.021)	0.018 (0.019)	0.028 * (0.017)
Income level	0.160 *** (0.048)	0.085 * (0.050)	0.042 *** (0.012)	0.019 * (0.011)	0.038 *** (0.011)
Education level	0.022 (0.027)	0.058 * (0.030)	0.006 (0.007)	0.013 * (0.007)	0.011 * (0.006)
Participation in community governance	0.565 *** (0.109)	0.953 *** (0.116)	0.148 *** (0.028)	0.217 *** (0.025)	0.214 *** (0.024)
Involvement in publicity and training	0.244 *** (0.098)	0.003 (0.108)	0.063 *** (0.026)	0.000 (0.024)	0.041 * (0.022)
Constant	−4.479 *** (0.330)	−4.007 *** (0.296)
athrho	0.110 ** (0.048)
rho	0.110 ** (0.047)
Obs	1983		1983	1983	1983

Note: *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Numbers in parentheses are robust standard errors.

presents detailed results, systematically organized into three key aspects: (1) direct effects of policy tools, (2) direct effects of policy perception, and (3) impacts of control variables. The correlation coefficient between the error terms of the two models was significant at the 5% level, indicating a substantial positive correlation between active and passive compliance behaviors. Marginal effects further clarify the practical significance and magnitude of each variable.

4.2.1. Direct Effects of Policy Tools

Supply-type tools significantly enhanced both active and passive policy compliance behaviors. Each additional unit of supply-type tools increased the probabilities of active and passive compliance by 0.4% and 0.5%, respectively, and raised simultaneous compliance by 0.6%, supporting hypotheses H1 and H1a. These findings corroborate the theoretical assertion that improved infrastructure, public services, and technical guidance substantially lower residents’ compliance barriers, fostering greater participation and compliance. Empirical evidence from Wuhan exemplifies this, where strategic investments in infrastructure—such as establishing extensive waste sorting points and optimizing collection schedules—have notably enhanced residents’ convenience, trust, and sustained compliance [61].

Environment-type tools had a significant positive impact predominantly on passive compliance. Each unit increase in these tools raised the probabilities of passive compliance by 0.3% and simultaneous compliance by 0.2%, confirming H1b. Aligned with social deterrence theory, environment-type tools such as strict legal regulations and robust supervision mechanisms effectively deter non-compliance by highlighting potential penalties and enforcement costs [62,63]. Case studies from Dalian illustrate this point, as monthly evaluations, public rankings, and mandatory corrective actions have substantially improved passive compliance.

Demand-type tools significantly promoted active compliance behaviors. Each unit increase in demand-type tools increased active compliance by 0.4%, supporting H1c. Consistent with externality theory, these tools internalize environmental benefits, motivating individual action through targeted incentives and participatory mechanisms, thereby reducing free-riding and promoting collective action [64]. Shenzhen’s public–private partnership model, employing advanced technology and management systems, exemplifies how effective demand-side incentives facilitate proactive compliance.

4.2.2. Direct Effects of Policy Perception

Policy cognition exerted a substantial positive effect on both compliance behaviors. Specifically, each unit increase in policy cognition corresponded to a 14.4% rise in active compliance, a 4.5% increase in passive compliance, and an 11.6% enhancement in simultaneous compliance. These results strongly support hypothesis H2 and underline the Theory of Planned Behavior’s assertion that greater policy knowledge strengthens intrinsic motivations, thereby boosting compliance [65]. Despite strong policy cognition, field studies highlighted challenges such as inconsistent waste collection management, suggesting the importance of aligning policy content awareness with reliable and efficient implementation practices [66].

Policy acceptance also showed significant positive impacts on both compliance behaviors. Each additional unit increase in policy acceptance raised active compliance by 4.4%, passive compliance by 8.2%, and simultaneous compliance by 7.3%, reinforcing hypothesis H2. Field insights from Suzhou underscore that when residents perceive policies as fair and legitimate, their normative and emotional support significantly boosts long-term compliance [67]. Suzhou’s integrated community engagement and targeted educational initiatives successfully cultivated broad societal support, indicating the critical role of acceptance in achieving sustainable policy adherence.

4.2.3. Effects of Control Variables

Several social–demographic and behavioral variables were significant predictors of compliance behaviors. Health status, income level, education level, community governance participation, and involvement in publicity and training positively influenced compliance. Conversely, age negatively impacted compliance, whereas gender, employment status, and housing status did not exhibit significant effects. These findings align with previous research indicating that younger, healthier, and more educated individuals with higher incomes and active community engagement tend to demonstrate higher compliance [68]. Identifying this key compliance group provides valuable insights for targeted behavioral interventions aimed at improving overall compliance with urban waste sorting policies.

4.3. Analysis of the Interaction Effects Between Policy Tools and Policy Perception

To explore the interaction effects between policy tools and policy perception in shaping residents’ compliance behaviors regarding municipal solid waste (MSW) sorting policies, we further introduced interaction terms into the bivariate probit model.

Table 3. Estimation results of interaction terms on residents’ compliance behavior with municipal solid waste sorting policies.

Variables	Biprobit (1)		Biprobit (2)		Biprobit (3)
	Active Policy Compliance	Passive Policy Compliance	Active Policy Compliance	Passive Policy Compliance	Active Policy Compliance	Passive Policy Compliance
Environment-type × Policy cognition	0.016 *** (0.004)	−0.006 * (0.004)
Environment-type × Policy acceptance	−0.011 *** (0.002)	0.008 ** (0.002)
Supply-type × Policy cognition			0.032 *** (0.007)	−0.006 (0.006)
Supply-type × Policy acceptance			−0.016 *** (0.004)	0.010 *** (0.004)
Demand-type × Policy cognition					0.032 *** (0.012)	−0.009 (0.009)
Demand-type × Policy acceptance					−0.032 *** (0.007)	0.018 *** (0.007)
Other variables	Control		Control		Control
athrho	0.127 *** (0.049)		0.119 ** (0.049)		0.123 *** (0.049)
rho	0.127 *** (0.048)		0.118 ** (0.048)		0.122 *** (0.048)
Obs	1983		1983		1983

Note: *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Numbers in parentheses are robust standard errors.

presents detailed empirical findings on these interaction effects, revealing complex moderating relationships that partially deviate from the initially proposed hypotheses H3 and H4.

First, policy cognition and acceptance significantly moderate the impact of environment-type tools on compliance behaviors. For residents with high policy cognition, environment-type tools enhance active compliance (β = 0.016, p < 0.01) but reduce passive compliance (β = −0.006, p < 0.1), suggesting that cognitively aware residents are more responsive to regulatory clarity but less susceptible to externally imposed routines. Conversely, among residents with high policy acceptance, environment-type tools are negatively associated with active compliance (β = −0.011, p < 0.01) yet positively associated with passive compliance (β = 0.008, p < 0.05). This divergent pattern implies that emotionally engaged residents may passively conform due to social or normative pressure while exhibiting resistance toward top–down enforcement perceived as overly restrictive. These findings underscore the need to differentiate between cognitive and emotional drivers when assessing the behavioral efficacy of environment-type tools.

Secondly, the interaction effects involving supply-type tools reveal distinct patterns. Among residents with high policy cognition, supply-type tools significantly enhance active compliance (β = 0.032, p < 0.01) but have no significant effect on passive compliance, suggesting that cognitively engaged individuals are more responsive to infrastructure-based facilitation. In contrast, for residents with high policy acceptance, supply-type tools are negatively associated with active compliance (β = −0.016, p < 0.01) and positively associated with passive compliance (β = 0.010, p < 0.01), indicating a potential motivational crowding-out effect. These findings underscore the differentiated roles of cognition and emotional acceptance in shaping responses to supply-side interventions and warrant further investigation in perception-targeted policy design.

Thirdly, demand-type tools demonstrated a comparable pattern of perceptual moderation. Among residents with high policy cognition, these tools significantly enhanced active compliance (β = 0.032, p < 0.01) but had no significant effect on passive compliance. This indicates that cognitively aware individuals are more likely to be motivated by participatory initiatives and incentive-based engagement to actively sort waste. Conversely, among residents with high policy acceptance, demand-type tools exerted a negative effect on active compliance (β = −0.032, p < 0.01) and a positive effect on passive compliance (β = 0.018, p < 0.01). This suggests that while emotionally aligned individuals may passively comply under external incentives, such tools could undermine their autonomous motivation to participate actively. These differentiated effects, particularly the reversal in active compliance, reinforce the need for perceptual targeting when applying demand-side instruments.

The significant and consistent correlation coefficients across interaction models (rho coefficients ranging from 0.118 to 0.127, all statistically significant at the 5% level or better) further underscore the interdependent nature of active and passive compliance behaviors.

These findings offer crucial implications for targeted policy interventions. They highlight the importance of matching specific policy tools with residents’ differing perceptions to maximize compliance effectiveness. Notably, cognitive-focused residents benefit substantially from clarity and transparency in policy implementation (environment-type tools), as well as practical facilitation (supply-type tools) and direct incentives (demand-type tools). Conversely, residents motivated primarily by emotional and normative acceptance require careful calibration of policy stringency and supportive mechanisms to prevent motivational crowding-out.

Overall, although the direct moderation hypotheses (H3 and H4) were not universally supported as initially proposed, the interaction results deepen the understanding of the interplay between policy tools and resident perceptions. These nuanced insights can guide policymakers in designing more precise and contextually appropriate interventions to foster sustainable urban waste management practices.

4.4. Analysis of the Interaction Effects Among Policy Tools

To further clarify the combined effectiveness of different policy tools, this study examined the interaction effects among the supply-type, demand-type, and environment-type policy tools. Given the insignificance of the correlation coefficient (rho) in the bivariate probit model, we employed a probit model to estimate the moderating effects among these tools. The detailed empirical results are presented in

Table 4. Moderating effect test considering the interaction of policy tools.

Variables	Probit (1)	Probit (2)	Probit (3)	Probit (4)	Probit (5)	Probit (6)
	Active Policy Compliance	Passive Policy Compliance	Active Policy Compliance	Passive Policy Compliance	Active Policy Compliance	Passive Policy Compliance
Environment-type	−0.020 *** (0.003)	0.014 *** (0.004)	−0.000 (0.003)	0.013 *** (0.004)	0.001 (0.003)	0.013 *** (0.004)
Supply-type	0.011 * (0.007)	0.023 *** (0.007)	0.004 (0.007)	0.021 *** (0.007)	0.006 (0.007)	0.021 *** (0.007)
Demand-type	0.098 *** (0.010)	−0.007 (0.009)	0.085 *** (0.010)	−0.003 (0.009)	0.059 *** (0.009)	−0.002 (0.008)
Environment-type × Supply-type	0.003 *** (0.000)	−0.000 (0.000)
Environment-type × Demand-type			0.005 *** (0.000)	0.000 (0.000)
Supply-type × Demand-type					0.008 *** (0.001)	0.000 (0.001)
Other variables	Control	Control	Control	Control	Control	Control
Obs	1983	1983	1983	1983	1983	1983

Note: * and *** denote statistical significance at the 10% and 1% levels, respectively. Numbers in parentheses are robust standard errors.

.

First, the interaction between environment-type and supply-type policy tools significantly enhances residents’ active compliance behavior (β = 0.003, p < 0.01). This indicates that robust regulatory environments, complemented by sufficient public infrastructure, financial subsidies, and technical support, substantially improve residents’ willingness to proactively adhere to MSW sorting policies. This finding aligns with the hypothesis (H1a, H1b) and underscores the synergy between environment-type and supply-type tools.

Second, the interaction between environment-type and demand-type policy tools also shows a significant positive impact on active policy compliance (β = 0.005, p < 0.01). Environment-type tools, such as clear regulations and rigorous supervision, effectively amplify the motivational effects of demand-side instruments, including educational initiatives, incentives, and participatory programs. This result provides further empirical support for integrating regulatory frameworks with participatory and motivational demand-side strategies—such as incentives, education, and resident engagement initiatives—to maximize policy compliance.

Third, the interaction between supply-type and demand-type policy tools similarly reveals a significantly positive effect on active compliance behavior (β = 0.008, p < 0.01). Effective public infrastructure and adequate resource allocation combined with incentives and education not only lower compliance barriers but also enhance intrinsic motivation among residents, promoting sustained active compliance behaviors.

However, none of these interaction terms significantly affected passive policy compliance outcomes. Specifically, the interaction coefficients for passive compliance in all three pairings (environment–supply, environment–demand, and supply–demand) were statistically insignificant. This suggests that the synergistic effects of policy tools primarily shape residents’ proactive behavioral responses rather than passive adherence, which may be more sensitive to top–down mandates or habitual factors. The findings align with theoretical arguments that active policy compliance is more dependent on comprehensive, integrated policy interventions, whereas passive policy compliance may not require as much strategic layering of tools.

Overall, these results highlight the importance of deploying complementary combinations of policy tools to achieve optimal compliance outcomes. In particular, the combination of environment-type and supply-type tools demonstrates the most robust and consistent effectiveness across both active and passive compliance behaviors. This suggests that well-designed regulatory frameworks, when matched with adequate infrastructure and service provision, create a reinforcing mechanism that motivates residents through both external enforcement and internal facilitation. Moreover, drawing on interaction effects and cross-comparison of tool performance, our synthesis identifies this combination as a cost-effective strategy characterized by relatively low capital expenditure (CAPEX) and medium-level operational expenditure (OPEX). These cost-efficiency patterns are derived from the functional characteristics of the tools involved—where infrastructure upgrades and standard regulatory enforcement typically require moderate investment and maintenance relative to demand-type strategies. This reinforces the utility of supply–environment combinations in resource-constrained urban contexts. These insights reinforce our theoretical propositions (H1a, H1b, H1c) by demonstrating that the effectiveness of MSW sorting policies hinges not only on the strategic combination of policy tools but also on the systemic alignment of financial, operational, and perceptual dimensions.

4.5. Discussion on Endogeneity and Robustness

To address potential endogeneity concerns—particularly those stemming from omitted variables or reverse causality between policy perception and policy compliance—this study adopted the Conditional Mixed Process (CMP) method based on a two-stage instrumental variable (IV) approach, as recommended in the prior literature [69,70]. Specifically, residents’ MSW sorting accuracy was selected as an instrumental variable, as it reflects cognitive understanding of policy content without directly influencing actual compliance behavior.

As shown in

Table 5. Endogeneity test results.

Variables	First-Stage Explained Variable: Policy Perception
	Coefficient	Robust Standard Errors	Coefficient	Robust Standard Errors
The accuracy of MSW sorting	3.485 ***	0.144	3.529 ***	0.145
Other variables	Control	Control	Control	Control
Constant	−2.070 ***	0.396	−2.316 ***	0.396
Variables	Second-Stage Explained Variable: Active Policy Compliance		Second-Stage Explained Variable: Passive Policy Compliance
	Coefficient	Robust Standard Errors	Coefficient	Robust Standard Errors
Policy perception	0.313 ***	0.089	0.499 ***	0.094
Other variables	Control	Control	Control	Control
Constant	−2.390 ***	0.244	−2.641 ***	0.262
atanhrho	0.630 ***	0.122	0.521 ***	0.118
LR statistic	1983.22 ***		2203.32 ***

Note: *** denote statistical significance at the 1% level. Numbers in parentheses are robust standard errors.

, the CMP estimation results provide strong evidence of endogeneity. The atanhrho values in both models were significantly positive at the 1% level, rejecting the null hypothesis of exogeneity. In the first-stage regression, MSW sorting accuracy showed a strong and statistically significant positive association with policy perception (coefficients = 3.485 and 3.529, both p < 0.01), confirming its strength and relevance as a valid instrument. In the second-stage regression, the estimated coefficients for policy perception on both active and passive compliance remained significantly positive (0.313 and 0.499, respectively; p < 0.01), aligning with the core findings of the bivariate probit model reported in Table 2. These results provide robust empirical support for Hypothesis 2, which posits a positive relationship between policy perception and compliance behavior.

To further validate the robustness of our results, we examined potential model misspecification by re-estimating models that incorporate interaction terms among policy tools (see Table 5). The results indicate that interaction effects between environment-type and supply-type tools, environment-type and demand-type tools, and supply-type and demand-type tools all have significantly positive impacts on residents’ active compliance at the 1% significance level. These findings confirm the complementary effects among policy tools and underscore the robustness of the main conclusions regarding their combined influence on policy compliance. This is consistent with previous studies that emphasize the synergistic effects of integrated policy tools in promoting pro-environmental behavior, particularly when tools are used in combination rather than isolation [71,72].

In sum, the endogeneity correction and robustness checks reinforce the reliability and stability of the study’s empirical findings. They highlight the importance of both instrumental factors (policy tools) and subjective drivers (policy perception) in shaping MSW sorting compliance, further validating the integrated analytical framework proposed in the theoretical section. These results provide credible evidence for designing more effective and nuanced policy interventions.

5. Discussion

This study investigates how policy tools and policy perception influence urban residents’ compliance with municipal solid waste (MSW) sorting policies in 34 Chinese cities. Drawing upon a dual theoretical framework—rational actor and behavioral models—we applied a bivariate probit model and conditional mixed process (CMP) estimations. Key findings indicate that both policy instruments and residents’ policy perceptions significantly affect active and passive compliance behaviors, with notable interaction effects between different types of policy tools and between policy perception and tool deployment.

Our results are consistent with the existing literature emphasizing the role of policy instruments in shaping pro-environmental behavior [73,74]. The finding that supply-type tools are particularly effective in promoting active compliance echoes prior studies on behavioral incentives in environmental governance. Additionally, the moderating role of policy perception aligns with Tyler [24] and Stern [44], who suggest that legitimacy and emotional acceptance enhance compliance.

However, contrary to expectations, policy acceptance, rather than cognition, exerts a stronger influence on passive compliance. This divergence from traditional “knowledge–attitude–practice” (KPA) models suggests that internalized emotional alignment may outweigh informational campaigns in some contexts, offering new insight into behavioral environmental policy [75].

Surprisingly, interaction terms reveal that environment-type tools significantly amplify the effectiveness of both supply-type and demand-type tools, suggesting a synergistic governance mechanism where regulatory clarity and normative structures create favorable conditions for material or incentive-based interventions. However, the effectiveness of such instruments—especially those promoting source separation—depends not only on behavioral responsiveness but also on the capacity of downstream infrastructure to sort and process additional waste fractions. Without sufficient collection systems, transfer stations, and material recovery facilities (MRFs), policy tools may fail to deliver expected outcomes.

Field observations further indicate that logistical issues such as overflowing bins, commingled collection, and irregular transfer operations erode public confidence and reduce residents’ perception of policy credibility. When individuals observe institutional lapses that contradict their sorting efforts, their motivation for sustained compliance weakens. These findings emphasize the co-dependence between perceptual trust and infrastructural adequacy, highlighting the need for integrated governance strategies that align front-end policy instruments with back-end service performance.

Our findings contribute to theory in two ways. First, they offer empirical support for a dual-perspective compliance model that incorporates both instrumental rationality and bounded rationality. Second, they reveal a cross-level interaction: macro-level tools (e.g., regulations) shape micro-level residents’ policy perceptions (e.g., cognition and acceptance), reinforcing the feedback loop proposed in the cognitive governance literature. These insights point toward the potential development of a dynamic, co-evolutionary model of environmental policy compliance.

Several limitations warrant acknowledgment. The cross-sectional nature of the data precludes analysis of temporal dynamics in perception and compliance behavior. Future studies could adopt longitudinal designs and incorporate qualitative insights to trace how policy trust, legitimacy, and participation evolve over time. Additionally, further research should explore the relative strength and cost-effectiveness of policy tools in varied urban governance contexts. Additionally, further research should explore the relative strength and cost-effectiveness of policy tools in varied urban governance contexts, including differences across housing typologies (e.g., high-rise vs. single-family dwellings) that may influence responsiveness to tariff structures or incentive programs.

6. Conclusions

This study examined how policy tools and policy perception jointly shape urban residents’ compliance with municipal solid waste (MSW) sorting policies, based on survey data from 34 Chinese cities and a bivariate probit modeling strategy. The findings reveal that environment-, supply-, and demand-type policy tools all exert significant positive effects on residents’ active and passive compliance, while policy perception—measured by both cognition and acceptance—plays an even more substantial role. Moreover, the interaction between policy tools and policy perception further amplifies their impact on compliance, highlighting the importance of a policy environment that is both well designed and well received.

These findings contribute to the literature by bridging policy instrument theory and behavioral compliance research within the context of urban environmental governance. In contrast to traditional knowledge–attitude–practice (KAP) models that emphasize individual-level determinants, this study demonstrates that institutional arrangements and residents’ subjective interpretations interactively influence policy effectiveness. The results thus enrich our understanding of how policy tools can be calibrated not only in type but also in communicative alignment with public perception to enhance implementation outcomes.

Moreover, the results highlight the importance of constructing coherent and communicative policy environments to foster sustainable behavioral change. Future waste governance efforts should move beyond siloed policy tools and adopt synergistic combinations of regulatory, infrastructural, and participatory approaches. Such an integrated and perception-sensitive strategy is essential for enhancing the legitimacy, responsiveness, and overall sustainability of MSW sorting systems in rapidly urbanizing contexts. Clean front-end sorting not only supports behavioral compliance but also enables low-carbon downstream valorization, such as gasification and green hydrogen production [76], reinforcing the systemic relevance of integrated MSW strategies.

Importantly, among the tested combinations, the integration of environment-type and supply-type tools demonstrated the most consistent effectiveness across both active and passive compliance behaviors. This pairing also exhibits a relatively favorable cost profile—characterized by moderate capital and operational expenditures—thereby offering a pragmatic and scalable strategy for cities operating under fiscal and infrastructural constraints.

Overall, this study underscores the need for an integrated policy approach—one that combines diverse policy tools with strategies to cultivate positive public perception—to advance the effectiveness and sustainability of urban MSW sorting systems.

7. Policy Implications

Drawing upon the empirical findings and theoretical insights, we offer four targeted policy recommendations to enhance urban residents’ compliance with municipal solid waste (MSW) sorting policies:

• Diversify and Contextualize Policy Tool Portfolios: Governments should adopt a balanced mix of environment-type, supply-type, and demand-type tools, avoiding overdependence on regulatory enforcement. Among these, the combination of environment-type and supply-type tools demonstrates the most robust and consistent effectiveness across both active and passive compliance behaviors while offering favorable CAPEX/OPEX profiles—making it especially suitable for resource-constrained urban contexts. Region-specific strategies are essential: regulatory instruments are more appropriate in less-developed areas, while incentive-based and participatory tools are better suited to regions with stronger institutional capacities.
• Align Policy Instruments with Public Perception: Policymakers should improve policy cognition through accessible and consistent information campaigns and foster acceptance via participatory mechanisms and trust-building measures. This alignment enhances both behavioral intention and emotional commitment.
• Tailor Tool Deployment to Perception Profiles: Policy tools should correspond to residents’ cognitive and emotional readiness. Coercive tools may deter behavior if not paired with supportive measures. Aligning tool types with perception characteristics helps avoid motivational crowding-out and strengthens voluntary engagement.
• Promote Adaptive and Feedback-Oriented Governance: Policymakers should institutionalize real-time feedback mechanisms—such as digital surveys and community consultations—to evaluate both tool effectiveness and public sentiment. Iterative adjustments based on these insights can sustain compliance and build public trust. Moreover, system-wide alignment across tool design, infrastructure capacity, and public perception should be prioritized to ensure coherent, resilient, and cost-effective implementation.