The Moderating Effect of Perceived Policy Effectiveness in Residents’ Waste Classification Intentions: A Study of Bengbu, China

Abstract

The Chinese government is promoting a waste classification policy to solve the increasingly serious issue of cities being besieged by waste. Only few studies investigate whether residents’ understanding of garbage classification policy has an impact on their garbage classification behaviour and the nature of such impact. The purposes of this study are twofold: first, to explore conceptually the mechanism behind any moderating effects of perceived policy effectiveness (PPE) on waste classification and, second, to examine empirically if and how PPE influences the relationships between attitude (ATT), subjective norm (SN), perceived behaviour control (PBC), awareness of consequence (AC) and waste classification intention (WCI). The conceptual model of the study is developed by combining insights from the theory of planned behaviour, norm activation theory and value–belief–norm theory. A total of 351 questionnaires were administered in person to households in Bengbu, China. The results based on structural equation modelling with partial least squares show that PPE negatively moderates the relationship between AC and WCI. AC is more strongly related with the intention to classify waste when PPE is weaker. Likewise, when PPE is higher, people’s awareness of consequences becomes less important for WCI. The findings have significant implications in policymakers’ developing guidelines and offer a framework for implementing more effective waste classification policy.

1. Introduction

Many Chinese cities suffer from severe environmental pollution and waste that is more extensive and serious than air pollution, such as fog and haze. According to the World Bank, China has surpassed the United States as the world’s largest waste producer, with an annual output of more than one billion tons. A report by the Ministry of Housing and Urban–Rural Development of China shows that two thirds of China’s cities are besieged by waste and one fourth of its cities have no space in which to build landfills.

Therefore, the widespread implementation of waste classification systems is crucial to improving the living environment of China’s nearly one billion people. China’s waste classification began in the 1990s. In June 2000, the Ministry of Construction designates eight cities across the country as the first batch of pilot cities for the classification and collection of domestic waste. In June 2016, the National Development and Reform Commission and the Ministry of Housing and Urban-Rural Development jointly issue the “Plan for the Mandatory Waste Classification System”, which proposes the mandatory classification of domestic waste as an important measure for green development and innovative urban management. On 2 June 2019, Chinese Chairman Xi Jinping issues important instructions on waste classification and clarified the extreme importance of waste classification for high-quality and sustainable development of the country; by this time, waste classification has been upgraded to a national strategy.

In effective waste classification, at the front-end, involving the residents is the key to mid-end classification, collection, and transportation, as well as back-end disposal and resource utilization. It is an important means of implementing the entire waste classification strategy and solving the dilemma of “waste siege” [1]. Residents’ awareness of and participation in garbage classification is fundamental to the success of waste classification [2,3].

Previous research has tested the validity of the theory of planned behaviour (TPB) [4] in waste classification or recycling behaviour [5,6,7]. These studies show the importance of public policies in residents’ behaviour [8,9]. Thereby, perceived policy effectiveness (PPE) is relevant to waste classification [10]. However, only few studies [8,11,12] investigate whether residents’ understanding of garbage classification policy has an impact on their garbage classification behaviour and the nature of such impact. The objective of this study, therefore, is to explore the mechanism through which PPE impacts residents’ waste classification behaviour.

We introduce a conceptual model, based on an extended TPB model, which then is used as a basis for investigating the moderating the effects of PPE on the relationships between attitude (ATT), subjective norm (SN), perceived behaviour control (PBC), awareness of consequence (AC) and waste classification intention (WCI). The empirical analysis is based on a sample of residents who have not yet implemented waste classification in a certain area of mainland China. Based on our findings we derive policy recommendations for a better understanding of the waste recycling intentions of residents.

2. Materials and Methods

2.1. Theoretical Background

The theory of planned behaviour (TPB model) is an extension of the theory of reasoned action (TRA Model) [13]. The TPB model is an important theoretical lens for studying public behaviour from a micro perspective. Environmental issues, such as water source protection [14], greenhouse gas emission [15], tourists’ environmental behaviour [9], sustainable transport behaviour have been studied using this model. As described by Ajzen [4], the flexibility in adding variables to the TPB model makes it a very useful tool for studying intention and behaviour. Many studies on the intention of waste classification have extended the original TPB model by adding relevant variables.

The basic TPB model mainly includes three main factors that predict individual behaviour: attitude (ATT), subjective norm (SN) and perceived behaviour control (PBC). Attitude is defined as the positive or negative evaluation or cognitive representation of an individual showing a particular behaviour. Subjective norm refers to the perception of the social pressures that individuals experience when they show a particular behaviour. For instance, in our application this means that residents are more willing to implement the behaviour when others who are important to them consider waste classification to be effective. Perceived behaviour control refers to the degree to which an individual is expected to control (or master) a given behaviour [4,16]. These three variables in the TPB model have also been identified as key predictors of waste classification intentions [7,11,17,18].

Norm activation theory (NAT model) assumes that personal norms are an enabler of moral behaviour. This theory focuses on internal determinants [19,20] such as the awareness of consequences. Awareness of consequences (AC), in this study, means that individuals can protect their living environment, reduce waste, reduce losses, and set a good example for future generations and their children through the implementation of waste classification. NAT models are commonly used in the exploration of citizens’ intention to protect the environment [3,7,11,21]. The more favourable are residents’ perceptions of the results from waste classification, the stronger their intentions to conduct waste classification will be [22]. Taken together, TPB and NAT suggest that individual intentions, in our study the intention to classify waste, crucially depend on four individual traits: ATT, SN, PBC and AC.

Value–belief–norm theory (VBN) suggests that, in addition to these four variables perceived Policy Effectiveness (PPE) matters [6,23]. Within VBN the PPE has been proven to have a good explanatory power for public environmental behaviour intentions such as classifying and recycling of waste [8,10]. In this study, PPE implies that residents understand the effectiveness of governmental waste classification policy. PPE has a direct impact on waste classification intentions (WCI) as an effective policy can act as an incentive for people to classify: If people perceive a policy to be effective, their intentions to classify waste will be increased [10].

2.2. Hypotheses about Moderating Effects of PPE

In addition to being a variable that directly affects intentions, PPE may also be a moderator for the other four variables that affect intentions to classify waste. Specifically, as detailed by [10], a low PPE (i.e., a perceived policy ineffectiveness) mutes the positive impact a positive attitude (ATT) has on the intention to classify waste.

If the people have a positive attitude toward recycling, their intention will decrease if there is no effective policy support [10]. Furthermore, if an individual’s attitude towards recycling is still poor under the premise of effective policies, her attitude may not change with the change of policy. Based on this discussion, we propose the following hypothesis:

Hypothesis 1 (H1).

The impact of attitude (ATT) on waste classification intention (WCI) is higher for a higher perceived policy effectiveness (PPE).

In the TPB model, subjective norm (SN) is an important factor affecting the intention to waste classify. In addition to the literature on the waste classification intention, SN is also treated as an important influencing factor in studies investigating environmental protection intention [1,14].

However, in the field of health psychology, it is proposed that if an individual has already formed a routine waste-classifying behaviour, the impact of SN on the WCI will be reduced accordingly [24]. This effect arises because the waste classifying behaviour has been internally solidified.

Before people have adequate knowledge about waste classification and develop a routine classifying behaviour, they will be greatly affected in their intentions to classify waste by third parties, such as their relatives and friends. In so far as an effective implementation of waste classification (that is a high PPE) shapes the perceptions of third parties about the importance of waste classification, a high PPE will also transmit via this channel to an individual’s intentions to classify waste. Specifically, in the process of continuously improving policies and people’s comprehensive knowledge about recycling, the impact of SN on recycling intention and behaviour will decrease accordingly [8]. Put differently, via public authority, a high PPE will aid the development of a routine waste classification behaviour, which, in turn, weakens the relationship between SN and WCI. Against this background, the following hypothesis is proposed:

Hypothesis 2 (H2).

The impact of social norm (SN) on waste classification intention (WCI) is lower for a higher perceived policy effectiveness (PPE).

Perceived behaviour control (PBC), as the motivation part of the ultimate classifying behaviour in the TPB model [25], requires people to acquire knowledge through autonomous or passive means to transform intentions into behaviour. In the process of passively acquiring waste classifying knowledge, the effectiveness of policies and social advocacy are of great significance. For example, individuals will enhance PBC when the government establishes convenient waste classifying facilities, such as the layout of residential waste classifying facilities and garbage sorting transport vehicles [26]. Ajzen [4] argues that PBC increases the motivation of individuals to perform a specific behaviour, which is perceived to be the ability to perform the behaviour. Thus, PBC and PPE can be seen as both intrinsic and extrinsic motivators [10]. When a person is intrinsically motivated to perform waste classification action (high PBC), WCI is higher. The extrinsic motivator PPE reinforces this link. Hence, the following hypothesis is proposed:

Hypothesis 3 (H3).

The impact of perceived behaviour control (PBC) on waste classification intention (WCI) is higher for a higher perceived policy effectiveness (PPE).

Awareness of consequences (AC) can be interpreted as people predicting future developments through existing policies and surroundings to decide whether to implement a certain behaviour [19]. Bamberg and Schmidt [27] suggest that people choose a behaviour that they think is better and easier to implement. External factors will influence the relationship between knowledge and an individual’s behaviour [28]. If there is a good social atmosphere to strongly support waste classification, the impact of individual perceptions such as active learning of waste classification will be reduced. When government policies can effectively stimulate WCI, the specifical awareness of waste classification consequence (AC) may not be necessary. Based on this, the following hypothesis is proposed:

Hypothesis 4 (H4).

The impact of perceived awareness of consequence (AC) on waste classification intention (WCI) is lower for a higher perceived policy effectiveness (PPE).

Figure 1 summarizes the conceptual background of this study:

3. Methodology

3.1. Measurement Instruments and Data Collection

To collect data, questionnaires were distributed (February 2021) to urban and rural residents in Bengbu, Anhui Province, China. Bengbu with its more than million inhabitants is an important industrial base in the Yangtze River Delta economic belt. It is also an economically developed city and a national comprehensive transportation hub in China. Bengbu is listed among the circular economy pilot cities of the National Development and Reform Commission in China. To date, Bengbu has not yet implemented compulsory waste classification.

The questionnaire as shown in Appendix A is developed based on existing literature [4,10,29,30]. The questionnaire is divided into two parts. The first part contains socioeconomic and demographic information, including gender, age, income, education level or area of residence. The second part aims to investigate residents’ intention to classify waste including the five latent variables: attitude (ATT), subjective norms (SN), perceived behaviour control (PBC), awareness of consequences (AC), and perceived policy effectiveness (PPE). A seven-point Likert scale is used to measure responses to questions. All variables range from “strongly disagree” (1) to “strongly agree” (7).

We applied the population proportion quota sampling method. To improve the effectiveness of the study, 30 questionnaires were distributed randomly before the official release. Based on the pilot test results, the questionnaire was modified. According to the population structure in Bengbu, we selected eight street communities for random convenience sampling. In China, the street community is the smallest unit of governance, and we dispatched the questionnaire through four graduate students and eight staff members of the street communities.

The questionnaires (400) were distributed from 1 to 31 February 2021. The effective questionnaire recovery rate is 87.75%. Details on the socio-economic and demographic background of respondents are included in

Table 1. Demographic and socioeconomic background of respondent.

Variable	Category	Frequency	Percentage (%)
gender	male	175	49.9
	female	176	50.1
Age	≤20	23	6.6
	21–35	157	44.7
	36–50	105	29.9
	51–65	46	13.1
	≥66	20	5.7
monthly income (in RMB)	≤5000	209	59.5
	5001–10,000	98	27.9
	10,001–15,000	26	7.4
	≥15,000	18	5.1
living area	urban	232	66.1
	rural	119	33.9

.

3.2. Data Analysis

Structural equation modeling (SEM) is a technique that includes a series of multivariate analysis methods such as regression analysis, factor analysis and analysis of variance. Partial least squares (PLS) is a common statistical method within SEM. It can be used to confirm the effectiveness of tool structures and to evaluate the structural relationship between these structures [31,32]. PLS requires less rigorous distributional assumptions, and it works well with nonnormal distributions and smaller sample sizes [31,33]. In this study, IBM SPSS17.0 and SmartPLS3.2.9 are used in combination with SEM.

The analysis includes two parts. The first part contains the measurement model (also called “outer model”) that displays the relationships between the constructs and the various indicator variables (i.e., items). The measurement model aims to evaluate the reliability and validity of the constructs derived from multiple items.

To evaluate the reliability of our constructs, two indicators are used: Cronbach’s alpha and composite reliability (CR). Cronbach’s alpha refers to a measure of internal consistency, which indicates how closely related the items comprising a group are. CR measures the sum of the latent variables’ factor loadings relative to the sum of the factor loadings plus error variance.

The validity analysis is divided into convergent validity and discriminant validity. Convergent validity refers to whether each indicator reflects the same construct. Average Variance Extracted (AVE) is an appropriate indicator for convergent validity [31]. Discriminant validity concerns the degree of correspondence between a certain structure and the measured value. Discriminant validity concerns the extent to which a construct is truly distinct from other constructs. In this study, we compare AVE and the correlation coefficient to check discriminant validity [31].

The second part of the analysis deals with the structural model that displays the relationships (paths) between the derived constructs [31]. The structural model aims to analyse the conceptual model’s ability to predict the variances of the dependent variable and the independent variables.

We apply a two-stage approach [34] to execute the statistical analysis. Stage one investigates the “main effects model”, that is the model without the interaction terms (Comparison Model). This model isolates the direct impact of the exogenous constructs (ATT, SN, PBC, AC and PPE) on the endogenous construct (WCI).

Stage two investigates the moderating effect of PPE. For this aim, the latent variable scores of each of the exogenous latent variables (ATT, SN, PBC, AC) and that of the moderator variable (PPE) are multiplied to create single-item measures, the interaction terms.

4. Results

4.1. The Measurement Model

Indicator variables (items) and their factor loadings are displayed in

Table 2. Reliability and Convergent Validity.

Variable	Item	Factor Loadings	Cronbach’s Alpha	CR	AVE
attitude (ATT)	ATT1	0.767
	ATT2	0.843	0.740	0.850	0.660
	ATT3	0.821
	SN1	0.859
subjective norm (SN)	SN2	0.873
	SN3	0.761	0.900	0.930	0.710
	SN4	0.890
	SN5	0.833
perceived behaviour control (PBC)	PBC1	0.845
	PBC2	0.820	0.820	0.880	0.650
	PBC3	0.774
	PBC4	0.775
awareness of consequence (AC)	AC1	0.822
	AC2	0.820	0.760	0.860	0.670
	AC3	0.816
perceived policy effectiveness (PPE)	PPE1	0.768
	PPE2	0.803
	PPE3	0.805	0.910	0.930	0.680
	PPE4	0.852
	PPE5	0.851
	PPE6	0.860
waste classification intention (WCI)	WCI1	0.850
	WCI2	0.696	0.780	0.840	0.580
	WCI3	0.614
	WCI4	0.851

. Minimum requirements for Cronbach’s alpha and composite reliability state that these measures should be greater than 0.7 [35,36]. As shown in Table 2, Cronbach’s alpha of all variables ranges between 0.74 to 0.91 and the CR values range from 0.84 to 0.93, suggesting that reliability is satisfactory.

The minimum factor loading and AVE requirements for each variable are that they are greater than 0.5 [35]. From Table 2 we see that the factor loadings of all the items range from 0.614 to 0.873 and the AVE values range from 0.58 to 0.71. Additionally, the factor loadings for all variables are significant at the one percent level and they are greater than 0.6.

Based on above results, the data are further analysed for discriminant validity. Cross-loadings are typically used to assess discriminant validity. The item’s/indicator’s (e.g., ATT1) outer loading on the associated construct (e.g., ATT) should be greater than any of its cross-loadings on other constructs [36,37]. As shown by

Table 3. Discriminant Validity.

	ATT	SN	PBC	WCI	AC	PPE
ATT1	0.767	0.547	0.558	0.487	0.547	0.432
ATT2	0.843	0.556	0.558	0.439	0.726	0.516
ATT3	0.821	0.521	0.514	0.374	0.729	0.489
SN1	0.574	0.859	0.578	0.444	0.542	0.575
SN2	0.553	0.873	0.645	0.563	0.561	0.561
SN3	0.619	0.761	0.650	0.481	0.619	0.526
SN4	0.595	0.890	0.607	0.560	0.626	0.609
SN5	0.505	0.833	0.560	0.554	0.496	0.539
PBC1	0.488	0.570	0.845	0.494	0.499	0.487
PBC2	0.556	0.601	0.820	0.482	0.587	0.522
PBC3	0.660	0.629	0.774	0.533	0.701	0.567
PBC4	0.460	0.512	0.775	0.543	0.462	0.484
WCI1	0.529	0.621	0.617	0.850	0.557	0.607
WCI2	0.263	0.302	0.350	0.696	0.292	0.308
WCI3	0.214	0.219	0.284	0.614	0.223	0.225
WCI4	0.505	0.572	0.568	0.851	0.504	0.562
AC1	0.701	0.519	0.542	0.401	0.822	0.439
AC2	0.722	0.545	0.607	0.403	0.820	0.510
AC3	0.605	0.578	0.574	0.548	0.816	0.545
PPE1	0.568	0.498	0.529	0.439	0.581	0.768
PPE2	0.539	0.546	0.567	0.454	0.567	0.803
PPE3	0.473	0.540	0.536	0.526	0.484	0.805
PPE4	0.471	0.557	0.544	0.507	0.490	0.852
PPE5	0.433	0.572	0.509	0.572	0.456	0.851
PPE6	0.462	0.574	0.503	0.517	0.489	0.860

, the component scores of each latent variable are higher than that for the other constructs, which indicates the model has good discriminant validity.

4.2. The Structural Model

The structural model is typically evaluated by calculating structural paths, t-statistics and the R2 value. We apply bootstrapping (5000 times) to calculate the statistical significance of the path coefficients.

Table 4. Estimation Results.

Constructs	Comparison Model			Conceptual Model
	Path Coefficients	t-Value	Sig.	Path Coefficients	t-Value	Sig.
ATT	0.025	0.309	0.757	0.018	0.220	0.826
SN	0.194	2.665	**	0.205	2.501	*
PBC	0.280	3.913	***	0.271	3.852	***
AC	0.061	0.789	0.433	0.041	0.482	0.630
PPE	0.253	3.977	***	0.267	4.055	***
ATT × PPE				0.148	1.728	0.084
SN × PPE				0.049	0.691	0.490
PBC × PPE				0.021	0.305	0.761
AC × PPE				−0.177	2.224	*
R2	0.502			0.510

Note: * p < 0.05; ** p < 0.01; *** p < 0.001; observations = 351.

summarizes the results for the main and moderating effects derived (see figure in Appendix B).

The two models demonstrate that about 50 percent of the variance in WCI can be explained by the predictors included. The R2 values range from 0.502 to 0.510. R2 values of 0.25, 0.50, and 0.75 indicate that the model validity is low, medium and high, respectively [37]. Hair et.al [38] propose a goodness of fit criterion (GoF) for PLS as the geometric mean of the average communality and R2. The values of 0.10, 0.25, and 0.36 could be interpreted as small, medium, or large effect [38]. The GoF of the model in this study is 0.57, implying that the model has a good fit.

In the comparison model, statistical significance is reached by variables SN, PBC and PPE. From the conceptual model we see that one of the interaction terms (i.e., AC × PPE) is statistically significant with a coefficient −0.177. These results imply that out of our four hypotheses only H4 finds empirical support: the effect of awareness of consequences on waste classification intention decreases with an increase in perceived policy effectiveness.

5. Discussion and Conclusions

This study contributes to the existing literature in three respects. First, it is one of the first papers to explore the psychological predictors of waste classification intentions in the rural and urban areas of China. Previous studies mainly deal with macro sociodemographic determinants within a single area [26,39].

Second, our study expands the previous research using the TPB model by integrating it with the NAT and the NAM models. Specifically, the paper includes variables “awareness of consequences” and “perceived policy effectiveness“ to analyse residents’ waste classification intentions based on data for Bengbu (China).Waste classification has not yet been implemented in Bengbu. The study empirically investigates moderating effects of perceived policy effectiveness in a waste classification intentions model. In this respect, this study responds to the call for more research on the TPB model [2,3,17,26,40]. The enhanced model is better equipped to help governments understanding the factors that affect residents’ waste classification intentions.

Third, the paper signals that perceived policy effectiveness has a negative impact on the link between the awareness of consequence and waste classification intentions. This finding contrasts with the conclusions drawn in previous studies [10,19]. The negative moderating effect is consistent with the view that by implementing effective policies and achieving good results governments can improve people’s intentions to adopt waste classifying behaviour [2,12]. This suggests that with the continuous improvement of policies the impact of awareness of consequence (AC) on waste classification intentions will become weaker, a view which has been voiced by Zhang et al. [2]. When people are highly motivated by policy measures such as the broad availability of classification facilities and by policy information via social media, AC has little impact on their intention to classify whether they are aware of the consequence.

Furthermore, regarding variable attitude (ATT), unlike previous studies [12,26,40], the paper finds that the direct impact of attitudes on waste classification intention is not statistically significant. A possible reason for this finding is that Bengbu is not yet a city that implements waste classification. According to free-talk with residents, we realized that their understanding of waste classification likely is not comprehensive enough. In Bengbu, there still exists a certain gap in the implementation of waste classification compared with cities like Shanghai, where the implementation has taken place and unsorted garbage disposal is deemed as illegal behaviour. This reveals the main challenges that governments may face in designing and implementing waste classification policies. In the process of policy implementation, the government should make policies and information open and transparent. If the perceived policy effectiveness increases, residents’ enthusiasm for waste classification will increase as well. Once people’s enthusiasm for waste classification has increased, and the uncertainty of the implementation of this policy has been gradually eliminated, people will be encouraged to effectively conduct waste classification [24,41].

What follows from our analysis for government policy? To strengthen the intention to classify waste, the policymakers also need to implement economic (dis-)incentives to encourage people to show a certain behaviour [12,14,42]. The government may also design an inducement policy or punishment mechanism to regulate the waste classification of residents [14,17]. Once residents develop good waste classification habits, the perception of the results of waste classification implementation will no longer be an important factor [10,12,42]. Furthermore, we suggest that governments continuously strengthen the construction of waste classification facilities to improve the convenience of residents’ waste classification [17]. The layout of the front-end classified delivery facilities, the effective operation of the middle-end classified transportation vehicles and the efficient operation of the dry and wet waste classifying and disposal facilities at the back-end can finally improve people’s sense of mission and responsibility [12,43]. Shanghai, as the most successful model of waste classification in China, has carried out the classification and treatment measures in the above three stages, and established an efficient and effective “point-station-hub” mode [44]. As of 5 January 2022, Shanghai has built 15,000 recyclables service points, 201 transfer stations and 10 distribution hubs, handling 14,782.51 tons of dry waste and 10,023.04 tons of wet waste per day.

The study has certain limitations that could be addressed by further research. The sample size of our study is relatively small (N = 351). Moreover, our results likely represent only the influencing factors of waste classification intention in Bengbu region and likely cannot be generalized. Specifically, the degree of waste classification implementation in the sampling area selected is low. Therefore, further studies could select two cities with different implementation levels of waste classification policy for comparison. This will help to gain additional insights into the effectiveness of perceived policies and recovery goals in different contexts across regions. Moreover, among the residential areas surveyed, three are adjacent to universities and the survey respondents are relatively young. To some extent, this could affect our results.

Despite these limitations, our findings regarding the factors shaping residents’ intention to classify waste should help governments to understand the main factors underlying waste classification intentions more fully. Our findings, thus, aid governments in their attempt to make new breakthroughs in policy design and implementation.