A factor analysis of residents’ performance in municipal solid waste source-separated collection: A case study of pilot cities in China

ABSTRACT

To explore the potential effectiveness of a localized waste recycling system in Chinese cities, several rounds of a municipal solid waste source-separated collection (MSWSC) program were implemented throughout China from 2000 to 2017. In our study, to evaluate the achievements of the MSWSC program, a structured questionnaire survey and face-to-face interviews were conducted in eight representative cities from the MSWSC pilot cities. Influencing factors were examined via a Hierarchical Linear Model (HLM) on both the individual level and the city level. The differences in MSWSC performance among the eight research cities were investigated through a comparative analysis. The results suggested that gender, age, knowledge of MSWSC, satisfaction with sanitation and satisfaction with publicity had significant impacts on MSWSC, and we further found that the city-level factors influenced the relationship between MSWSC performance and the individual-level variables. Differences among the eight research sites indicated that Nanjing and Hangzhou exhibited relatively high achievements in the implementation of MSWSC program, while Yichun and Guiyang showed poor performance. The experiences of these advanced pilot cities suggest that specific containers for designated waste types, widespread public education, and sufficient MSWSC incentives must be provided and publicized to promote municipal solid waste recycling behaviors. Given the unique status of municipal solid waste management in China and its regional heterogeneity, the MSWSC system should be further discussed with regard to local conditions in a wider range of city settings.

Implications: This study attempts to discuss the factors affecting performance on municipal solid waste source-separated collection (MSWSC) by considering the nested data of two levels of individual and city from a structured questionnaire survey. Therefore, a Hierarchical Linear Model is established which can analyze the intra-group and inter-group effects of two-level variables. The proposed method can be expanded to other cities to determine the main factors affecting MSWSC or to evaluate the residents’ performance on MSWSC.

Introduction

As the world largest municipal solid waste (MSW) generator, China has produced approximately 179 million tons of MSW in 2015 (Chen, Geng, and Tsuyoshi 2010; Guo et al. 2017; Shekdar 2009; Tai et al. 2011). Its total amount of MSW was only 31.3 million tons in 1980, but it was projected to reach at least 480 million tons in 2030 (Zhang et al. 2015; Zhang, Tan, and Gersberg 2010). This astonishingly increasing rate of MSW generation has accompanied the fast-growing population and economy on one hand (Fei et al. 2016; Linzner and Salhofer 2014), and the shortage of resources and technologies for MSW treatment on the other hand (ISWA and UNEP, 2002). As a consequence, China has been suffering from serious pollution triggered by MSW (Dong, Tong, and Wu 2001; Zheng et al. 2014).

Within this context, Environmental Protection Law of the People’s Republic of China, enacted on December 26, 1989, firstly established a legal framework for waste reduction including MSW (Standing Committee of the National People’s Congress, 1989). From then on, municipal solid waste management (MSWM) has continued to improve with the implementation of new regulations and policies (Chen, Geng, and Tsuyoshi 2010). Among the policies and regulations that were implemented to facilitate the MSWM, a significant attempt was the establishment of eight pilot cities, including Beijing, Shanghai, Guangzhou, Shenzhen, Hangzhou, Nanjing, Xiamen and Guilin, for municipal solid waste source-separated collection (MSWSC) in June 2000 (Zhang et al. 2012). MSWSC means residents’ classification of MSW into two or more categories before the MSW is further treated (Zhang and Wen 2014). This method has proven to be an effective pre-treatment for MSW before disposal in landfills or incinerators, given that MSWSC not only reduces the quantity of MSW but also recovers some raw materials (Gaeta-Bernardi and Parente 2016; Izagirre-Olaizola, Fernández-Sainz, and Vicente-Molina 2015). In order to promote and expedite MSWSC throughout China, central and local governments have promulgated targeted policies to standardize the MSWSC system and publicize advanced cities’ experience with the MSWSC (Fei et al. 2016). In 2015, another 18 cities were added to the list of pilot cities for MSWSC (MOHURD, 2015). According to the State Council of China (2017), at least 46 more cities will enforce the pilot MSWSC program by the end of 2020.

Despite these efforts, MSWSC performance in China still lags behind that of many other developed countries (Chung and Lo 2004; Zhang, Tan, and Gersberg 2010). In particular, the MSWSC participation and accuracy rates are unsatisfactory in China (Han and Zhang 2017; Wang and Nie 2001) and are further enfeebled by the lack of well-defined MSW classifications (Yu 2009). In addition, there are inequalities in different cities’ MSWM, and progress in legislation and policies regarding MSWM varies across the country (Chen, Geng, and Tsuyoshi 2010). Only a few cities, such as Beijing, Shanghai and Shenzhen, have formulated local regulations on the MSWSC (Shanghai People's Government, 2011; Shenzhen People’s Government, China 2015; Standing Committee of Beijing Municipal People’s Congress, 2011), while most remaining cities inform residents about MSWSC via notifications rather than legislation (Yu 2009). Furthermore, MSWSC enforcement is also undermined by the lack of public knowledge, concerns, and beliefs about MSWSC (Chung and Lo 2004; Yuan et al. 2006). Residents who are not equipped with knowledge about MSWSC may not be able to properly sort the MSW (Alexander et al. 2009). However, people who are familiar with and supportive of recycling may not necessarily be able to effectively transfer their knowledge and awareness into action (Borgstede and Biel 2002).

In this sense, many studies have sought to investigate factors that influence residents’ performance in MSWSC, aiming to reduce pollution through behavioral control of littering (Al-Khatib et al. 2009). Many analytical methods have been introduced into this field from different perspectives. For example, Chen, Li, and Ma (2015) adopted a binary logistic regression model to analyze the effect of individual-level variables on MSWSC, and further explored its internal mechanism. Echegaray and Hansstein (2016) adopted the model of Theory of Planned Behavior (TPB) to study the difference between e-waste classification behaviors and residents’ intentions in large and medium-sized cities in Brazil, and to explain the main constraints of MSWSC from a psychological perspective. Seacat and Northrup (2010) introduced Information Motivation Behavior (IBM) model into the field of public health behavior to explore the factors influencing MSWSC in British community. Most of these studies have examined the relationship between MSWSC and individual-level influencing variables, such as gender, age, education and income (Kofoworola 2007; Pakpour et al. 2014; Purcell and Magette 2010; Rahardyan et al. 2004; Saphores et al. 2006).

However, few studies have considered city-level variables influencing MSWSC (Al-Khatib et al. 2015). Therefore, it is necessary to study MSWSC through multilevel analysis to explore how city-level and individual-level factors influence MSWSC together (Kollmuss and Agyeman 2002; Steg and Gifford 2005). In order to address the gap in research on the MSWSC, this study attempts to discuss the factors affecting the MSWSC performance by analyzing the data of individual level and city level. Therefore, a Hierarchical Linear Model (HLM) is established, which has the ability to analyze the intra-group and inter-group effects of two levels variables as well. More frequently cited examples of HLM occur in sociology and economics, as data from these subjects often have a hierarchical structure. Simultaneously the HLM is also suitable for this study, as it regards individuals and groups as parts of a whole, and recognizes that MSWSC is conceptualized as the result of interactions between two levels variables. In addition, according to Cohen (1988), we verified the validity of the model by calculating the within-group difference and between-group difference (or Intra-Class Correlation, ICC). These facts indicate that the HLM model is feasible and effective.

The rest of this study is arranged as follows: Section 2 describes the data collection and research methods. Section 3 analyses and discusses the results of the model. Section 4 concludes the paper with the main conclusions and some policy suggestions. The main objectives of this work are as follows: (1) to examine the social characteristics of the respondents, including gender, age, income, occupation, and education level, and residents’ awareness regarding MSWSC issues; (2) to conduct a comparative analysis of the differences in MSWM among the research sites; (3) to identify factors relevant to the MSWSC program and establish an index system to evaluate MSW status; and (4) to provide insight for decision makers to use in managing MSW activities from the perspective of not only the individual level but also the city level.

Methodology

Study area

To investigate the MSWSC performance, a national survey was conducted in eight pilot cities throughout China. These cities were selected based on the following considerations. First, the selected cities should be spatially independent to avoid spatial homogeneity, and accordingly, the eight selected cities are geographically distributed throughout Northern, Eastern, Central, Southern and Southwestern China (Figure 1). Second, the selected pilot cities should feature a range of time spans for program implementation in order to consider performance over time. Therefore, we selected these cities adopting the program before 2000 (such as Beijing and Guangzhou), after 2010 (such as Yichun and Guiyang), and between the two years (such as Chongqing and Kunming).

Figure 1. The map of the study area.

Model method

Many studies have argued that it is necessary to study the effects of individual and contextual variables on environmental behaviors (Gelissen 2007). To explore which factors affect residents’ source separation behavior at the individual and city levels, the model of HLM is adopted. When data exist within two levels, the regression equation is built by using the variables in the first level. Then the intercept and slope of the above equation are regarded as dependent variables and the variables in the second level are taken as independent variables, and new equations are built. We can explore the influence of variables at different levels on the dependent variables through the model (Hofmann 1997). Since the intercept and slope in the first level regression equation are used as dependent variables in the second level regression equation, this practice is also called the regression of regression.

We hypothesized that at the individual level, age, gender, and other variables would be positively associated with MSWSC. At the city level, investment in sanitation, rewards for MSWSC and other variables were predicted to be associated with MSWSC. When variance decomposition is conducted, the traditional linear model cannot discriminate the group effects of different levels of data because doing so would increase the error term of the model (Gao et al. 2017). In consideration of this limitation, this study created individual-level and city-level data to explain the variables affecting participation in MSWSC. Additionally, we considered cross-levels influences. The application of the HLM which is consist of three models, including a null model, a random-effects regression model, and a full model.

Null model

The null model includes no independent variables in its equations. The null model focuses only on the differences in percentages of the dependent variable caused by the individual-level or city-level factors. According to Cohen (1988), it is necessary to establish the HLM model to evaluate the effect of the variables at each level when the between-group difference exceeds 5.90%. The equations are as follows:

Level-1 Model:

(1) $Y_{ij}=B_{0j}+R_{ij}$(1)

Level-2 Model:

(2) $B_{oj}=G_{00}+U_{0j}$(2)

(3) ${\rho }_1={\sigma }^2/\left({\sigma }^2+{\tau }_{00}\right)$(3)

(4) ${\rho }_2={\tau }_{00}/\left({\sigma }^2+{\tau }_{00}\right)$(4)

where j corresponds to the city; i corresponds to the individual; Y_ij is the dependent variable, which is the performance in the MSWSC of respondent i in city j; the numbers, such as 0, 1 and 2, are diacritical marks; B_0j is the regression coefficient of the individual-level variable to the independent variable; R_ij is the random element of the individual level; G₀₀ is the intercept of the city-level variable; U_0j is the random element of the city level; ρ₁ is the within-group difference; ρ₂ is the between-group difference; σ² is the variance at the individual level; and τ₀₀ is the variance at the city level.

Random effects regression model

The independent variables at the individual level are introduced into a random-effects regression model in order to determine whether it can be used to establish the city-level equation. This model is established as follows:

Level-1 Model:

(5) $Y_{ij}=B_{0j}+B_{1j}{X}_{ij}+R_{ij}$(5)

Level-2 Model:

(6) $B_{oj}=G_{00}+U_{0j}$(6)

(7) $B_{1j}=G_{10}+U_{1j}$(7)

where X_ij is the independent variable of level-1; B_1j is the slope of X_ij; G₁₀ is the intercept of the city-level variable to B_1j; and U_1j is the random term of the city level. The other variables are the same as those in formulae (1) and (2).

Full model

If the individual-level independent variables show significance in a test of random effects in the previous model, we must study the variation in the regression coefficients of these variables in different cities. Therefore, the city-level variables should be added into the full model:

Level-1 Model:

(8) $Y_{ij}=B_{0j}+B_{1j}{X}_{ij}+R_{ij}$(8)

Level-2 Model:

(9) $B_{oj}=G_{00}+G_{01}{W}_j+U_{0j}$(9)

(10) $B_{1j}=G_{10}+G_{11}{W}_j+U_{1j}$(10)

where W_j is the independent variable at the city level; G₀₁ is the slope of the city-level variable; and G₁₁ is the slope of the city-level variable. The other variables are same as those in formulae (1), (2), (5), (6) and (7) (Hofmann 1997).

We used the HLM 6 software for data processing and other operations, which was developed by the company Scientific Software International. It takes the hierarchical structure into account, and estimates of standard errors are obtained. Additionally, valid tests and confidence intervals can also be established, and variables at different levels used in the sample can be introduced into the model by HLM 6 (Scientific Software International, America 2018).

Variable design

The name, symbol, and definition of the selected variables in this study is shown in Table 1. More explanation for these variables follows.

Table 1. Descriptions of the selected variables.

	Item	Symbol	Definition
Independent variables	Performance on MSWSC	PM	The ratio of the actual number of MSW items with correct separation to 27
Level-1 (individual level)	Gender	GE	Male = 1; female = 2
	Age	AG	Under 18 = 1; 18 ~ 25 = 2; 26 ~ 35 = 3; 36 ~ 45 = 4; 46 ~ 55 = 5; over 56 = 6
	Education level	EL	Junior high school or lower = 1; senior high school = 2; junior high school = 3; undergraduate = 4; graduate or above = 5
	Family monthly income	FM	0 ~ 3,000 = 1; 3,000 ~ 6,000 = 2; 6,000 ~ 10,000 = 3; 10,000 ~ 16,000 = 4; 16,000 or above = 5
	Environmentalism	EN	Whether the respondent or respondent’s family members are environmentalists (respondent = 1; respondent’s family members = 2; neither = 3)
	Knowledge of MSW category	KM	Whether respondent knows about the MSW category in the research city (yes = 1; no = 2)
	Accuracy of MSWSC in simulation	AM	The ratio of the number of MSW items with correct answer in simulation to 27
	Satisfaction with sanitation	SS	Whether the respondent is satisfied with sanitation (yes = 1; no = 2)
	Satisfaction with MSWSC	SM	Whether the respondent is satisfied with MSWSC (yes = 1; no = 2)
	Satisfaction with publicity	SP	Whether the respondent is satisfied with publicity (yes = 1; no = 2)
Level-2 (city level)	Per capita gross domestic product	PC	Per capita gross domestic product in the research city (thousand RMB)
	Investment in sanitation	IS	The ratio of investment in sanitation to fiscal expenditure
	Reward for MSWSC	RM	The ratio of respondents rewarded for MSWSC to total respondents
	Duration of MSWSC	DM	The duration of MSWSC program

Socio-economic characteristics

Many studies identified socio-economic factors, including age (AG), gender (GE), education level (EL), family monthly income (FM) and other factors related to MSWSC (Ishitani 2010; Scannell and Gifford 2010). We hypothesize that senior citizens, women, people with high education levels and environmentalists (EN) are more inclined to correctly separate MSW. According to Ishitani (2010), age, educational level, family monthly income, environmentalism, etc., are aggregated as categorical variables in this study.

Knowledge of MSWSC

Knowledge is considered an important factor influencing MSWSC (Hornik et al. 1995; Welfens, Nordmann, and Seibt 2016). We hypothesize that respondents’ knowledge of the MSW categories could promote MSWSC. We investigated respondents’ knowledge of MSWSC from two perspectives: knowledge of MSW categories (KM) and accuracy of MSWSC in simulation (AM).

In order to assess whether respondents know about the MSW categories in their city, we investigate the official categories for MSW separation in each research site (Table 2). The eight cities can be divided into two groups. In Beijing (Dai 2016), Nanjing (Ding et al. 2015), Hangzhou (Long and Shen 2017), Chongqing (CMPG (Chongqing Municipal People’s Government), China 2016) and Guangzhou (GMPG (Guangzhou Municipal People’s Government), China 2015), MSW is separated into four categories: recyclables, hazardous waste, kitchen waste and other waste. The remaining cities, including Yichun (JPPG (Jiangxi Province People’s Government), China 2017), Guiyang (GCAB (Guiyang city administration bureau), China 2016) and Kunming (KMPG (Kunming Municipal People’s Government), China 2012), encouraged residents to classify MSW into recyclables, hazardous waste and other waste.

Table 2. Current MSW sorting method in the research cities.

City	Recyclables	Hazardous waste	Kitchen waste	Other waste/Non-recyclables
Beijing	√	√	√	√
Nanjing	√	√	√	√
Hangzhou	√	√	√	√
Chongqing	√	√	√	√
Yichun	√	√		√
Guiyang	√	√		√
Kunming	√	√		√
Guangzhou	√	√	√	√

Concerning AM, evidence showed that residents’ practice of MSWSC could be facilitated by accessibility to the opportunities to separating MSW and corresponding knowledge of MSWSC (Keramitsoglou and Tsagarakis 2013; Refsgaard and Magnussen 2009). Therefore, we assessed respondents’ knowledge of MSWSC by asking them to separate 27 items of MSW into four types, including recyclable waste, hazardous waste, kitchen waste, and other waste. Then, we calculated the ratio of the number of MSW items with correct answers in simulation to 27.

Concerns about MSWSC

Following previous studies (Zeng et al. 2016; Zhang et al. 2012), we investigated the barriers to respondents’ satisfaction with the MSWSC program from three perspectives: MSW facilities (A), neighbor awareness and behavior (B) and MSWM (C). Three questions were used to assess whether each factor was the major obstacle to implementing the MSWSC program. The respondents were required to respond 0 for “good” and 1 for “limited”, and we calculated the ratio of the score to the number of valid questionnaires in each city for each factor; therefore, a higher percentage indicated stronger dissatisfaction with the MSW category. We hypothesize that residents who are highly satisfied with the MSW category correctly separate more MSW than do residents who are less satisfied with the MSW category.

City-level variables

MSWSC generally focusses on a single level of analysis (Watson, Chemers, and Preiser 2001), but we try to hypothesize that city-level variables mediate the relationship between individual-level variables and MSWSC from four perspectives: higher per capita gross domestic product (PC), higher investment in sanitation (IS), more reward for MSWSC (RM) and longer duration of MSWSC (DM) will promote MSWSC. Detailed information about these variables in the research cities are presented in Table 3 (NBSC (National Bureau of Statistics of China), China 2016).

Table 3. Information about the MSWSC in the research cities.

City	Per capita gross domestic product (thousands RMB)	Investment in sanitation (millions RMB)	Fiscal expenditure (million RMB)	Duration of MSWSC (years)
Beijing	106.5	1911	45,246.69	20
Nanjing	118.2	2684	4212.05	16
Hangzhou	139.7	545	9611.77	16
Chongqing	52.3	162	3304.33	7
Yichun	27.8	214	3048.68	1
Guiyang	63.0	220	4486.30	6
Kunming	59.7	46	5936.56	11
Guangzhou	136.2	1221	14,362.23	24

Performance on MSWSC

Relatively high rates of knowledge and concern do not necessarily guarantee satisfactory behavior (De and De 2010). Therefore, we investigated the popularity of source separation in eight cities, and our conclusion accords with the previous findings that most Chinese citizens are willing to separate MSW (Chung and Poon 2001; Liu 2003), even though some interviewees declined to sort any items in our study. With this result in mind, we further explored the performance on MSWSC (PM, the ratio of the actual number of MSW items with correct separation to 27). Compared with the knowledge of MSWSC, the performance on MSWSC, on the basis of accurate understanding, focuses on whether respondents correctly separate MSW in their daily lives.

Data collection

Individual-level factors were obtained from our questionnaire, and city-level variables were derived from the statistical yearbook for each city (except reward for MSWSC). First, the public’s socio-economic characteristics, knowledge, and awareness of MSWSC were used as independent variables that affected MSWSC. These variables were obtained through the questionnaire survey and used as the first-level factors. Second, information about MSWSC in eight cities was collected from the NBSC (NBSC (National Bureau of Statistics of China), China 2016) to facilitate further analysis and promote a discussion of the effect of city-level factors on MSWSC (Table 3).

The data of the above-mentioned individual-level factors and the reward for MSWSC used in our study were acquired from questionnaire investigations in the study area from May to September 2016. As the eight pilot cities promoted MSWSC in some pilot communities in the urban area rather than the whole city, the respondents were randomly selected among residents living in these pilot communities implementing the MSWSC program. Face-to-face interviews were adopted in the questionnaire survey, and almost all the information was elicited from the designed questions. Additionally, we designed a team of 10 well-trained students from East China Normal University to conduct the field interviews. The questionnaire included three sections: Section 1 asked about the respondents’ knowledge of the MSW categories, their satisfaction with it, and incentives for MSWSC. In section 2, the respondents were asked to separate their MSW so that we could estimate how MSWSC was implemented in these cities. Section 3 requested information on the respondents’ age, income, and other socio-economic characteristics. Because a large sample size is necessary for this type of nationwide survey, over 100 questionnaires were distributed in each of the eight research cities. In total, 800 valid questionnaires were received from 832 distributed questionnaires, for a high response rate of 96.15%. A summary of the questionnaire collection in the eight cities is shown in Table 4.

Table 4. Summary of questionnaire collection in the research cities.

City	Distributed questionnaires	Valid questionnaires	Response rate (%)
Beijing	102	98	96.08
Nanjing	105	102	97.14
Hangzhou	106	103	97.16
Chongqing	103	100	97.09
Yichun	103	99	96.12
Guiyang	104	100	96.15
Kunming	105	101	96.19
Guangzhou	104	97	93.27

Results and discussion

Socio-economic characteristics

Table 5 shows the respondents’ socioeconomic characteristics. Approximately 40% of all respondents were males, and the remaining 60% were females. The highest response rate was found in the age groups of 26–35 (24.75%) and over 56 years old (23.63%), while the response rate of young people aged 18 and under was the lowest (4.00%). Additionally, 22.25% of the respondents or their relatives were engaged in environmental protection activities. More than half of the respondents had obtained a bachelor’s degree or higher. Regarding their income levels, approximately 80% of the respondents had a family income of between RMB 3,000 and RMB 16,000, while the response rates of interviewees whose family income was less than RMB 3,000 (7.6%) or over RMB 16,000 (13.5%) were the lowest.

Table 5. Socio-economic characteristics of the respondents.

Factor	Response	Count	Percentage (%)
Gender	Male	320	40.00
	Female	480	60.00
Age (years)	Under 18	32	4.00
	18–25	126	15.75
	26–35	198	24.75
	36–45	142	17.75
	46–55	113	14.13
	Over 56	189	23.63
Education level	Junior high school or lower	178	21.50
	Senior high school	182	22.76
	Junior high school	153	19.13
	Undergraduate	239	29.88
	Graduate or above	54	6.75
Family monthly income (RMB)	3,000 or lower	61	7.63
	3,000–6,000	210	26.25
	6,000–10,000	266	33.25
	10,000–16,000	155	19.38
	16,000 or above	108	13.50
Environmentalism	Respondents	80	10.00
	Relatives	98	12.25
	Neither	622	77.75

Knowledge of MSWSC

The accuracy of respondents’ choices of categories for MSW in the local MSW categories was approximately 54.0%. Nanjing achieved the highest accuracy (70.6%), and only the accuracy rates in Yichun and Guiyang were less than half: at 46.5% and 40.0%, respectively (Figure 2). Overall, most respondents understood which kind of MSW are classified according to their city request. These findings, which relate to the progress of MSWSC in the research sites, indicated that the respondents in the earliest pilot cities, Beijing, Nanjing, Hangzhou, Chongqing, Kunming and Guangzhou, were more likely to separate MSW into the correct trashcans, while interviewees in Guiyang and Yichun (the two cities with the shortest time since implementing the MSWSC program, Table 3) may have been less likely to classify MSW properly due to their low rate of awareness about MSWSC. Geographically, the collection efficiency is better in eastern China, such as in Nanjing and Hangzhou, than in other regions, as the environmental awareness and the infrastructure for environmental protection in other regions are more limited than the relatively developed eastern cities (Zhang, Tan, and Gersberg 2010).

Figure 2. Knowledge of the MSW categories.

Accuracy of MSWSC in simulation

The eight cities respondents’ total sorting accuracy in the four-category MSWSC system was 62.3%, and the accuracy rates (the ratio of the number of MSW items with correct answers in simulation to 27) for recyclable waste, hazardous waste, kitchen waste, and other waste were 68.2%, 64.1%, 69.3% and 65.7%, respectively (Table 6). Additionally, there was no significant difference between the accuracies of MSWSC simulation for each MSW type across research regions, especially for recyclable waste and hazardous waste (approximately 60–70%) (Table 7). We found that the respondents were able to correctly separate the majority of the MSW but were confused about the categories of several items, such as electronic waste, hazardous waste, and some special waste. According to Yu et al. (2010), peddlers and informal collectors offer more attractive incentives to consumers for electronic waste (such as U disks and DVDs) than does the official recycling system; thus, residents usually sell these MSW items to collectors and the recycling opportunity is lost. Another potential reason for poor public MSWSC participation is a lack of a unified regulation for the classification of electronic waste (Gu et al. 2016). Additionally, the relatively low rate of awareness of the differences between non-hazardous and hazardous waste, especially for modern energy-saving lamps and X-ray films, leaves local authorities with the challenge of educating residents and disseminating knowledge of MSWSC and its practices.

Table 6. Total accuracy of the MSWSC simulation for each MSW type.

Item		Type (%)
		Recyclables	Hazardous waste	Kitchen waste	Other waste
Recyclables	1 Clothing	89.75	0.50	0.00	9.75
	6 Toys	71.50	9.38	1.25	17.88
	10 DVDs	38.50	41.75	1.25	18.50
	12 Shoes	69.63	7.75	0.88	21.75
	13 U disks	35.25	45.00	0.63	19.13
	19 Newspapers	95.75	0.75	1.25	2.25
	20 Edible oil drums	80.13	4.63	9.88	5.38
	21 Medicine bottles	47.75	30.13	1.50	20.63
	22 Cans	93.63	2.38	1.13	2.88
	25 Packing boxes	59.63	11.00	13.25	16.13
	Mean	68.15	15.33	3.10	13.43
Hazardous waste	4 Energy-saving lamps	18.25	54.63	2.25	24.88
	14 Expired medicine	4.50	75.63	2.25	17.63
	15 Paint buckets	18.50	70.38	0.75	10.38
	23 X-ray film	25.38	38.88	1.13	34.63
	26 Rechargeable batteries	12.38	80.88	1.13	5.63
	Mean	15.80	64.08	1.50	18.63
Kitchen waste	2 Fruits and vegetables	3.75	2.13	89.25	4.88
	7 Mushrooms and jujubes	4.13	3.63	84.13	8.13
	8 Chinese medicine and tea	4.25	9.88	60.75	25.13
	11 Flour	5.50	3.13	81.00	10.38
	17 Chili paste	1.88	3.00	84.38	10.75
	18 Expired snacks	2.75	24.13	61.88	11.25
	24 Flowers	11.75	2.50	21.88	63.88
	27 Expired pork	1.50	20.38	70.88	7.25
	Mean	4.44	8.60	69.27	17.71
Other waste	3 Broken dishes	14.63	15.38	8.88	61.13
	5 Dust	8.25	12.75	3.13	75.88
	9 Toilet paper	21.38	17.38	5.50	55.75
	16 Diapers	9.50	15.63	5.00	69.88
	Mean	13.44	15.29	5.63	65.66

Table 7. Accuracy of the MSWSC simulation for each MSW type in each city.

City	Recyclable waste	Hazardous waste	Kitchen waste	Other waste	Total
Beijing	69.49%	65.31%	74.49%	63.27%	69.27%
Nanjing	68.33%	65.69%	75.00%	70.59%	70.15%
Hangzhou	65.24%	63.69%	77.18%	68.93%	69.03%
Chongqing	69.30%	67.60%	73.38%	63.00%	69.26%
Yichun	69.60%	64.04%	60.48%	58.84%	64.27%
Guiyang	66.40%	60.20%	62.25%	61.00%	63.22%
Kunming	68.91%	64.36%	58.54%	61.63%	63.92%
Guangzhou	68.04%	61.64%	72.68%	78.09%	69.72%

Concerns about MSWSC

First, as shown in Figure 3, residents of the eight cities considered that poor neighbors’ performance in MSWSC (36.1%) was the major barrier to implementing the program. The highest selection rate (66.5%) among the nine questions was the lack of awareness of MSWSC (B3), which, from the respondents’ perspective, may have a major influence on residents’ participation. Moreover, an unsatisfactory MSWM (31.6%) was the second-highest-rated obstacle to MSWSC, and 41.4% of all respondents suggested that the local government should take an active role in improving MSWM via well-designed educational and publicity programs (C1). Finally, the low score for dissatisfaction with the MSW facilities (26.3%) indicated a relatively good performance of this factor, especially for the “type of trashcans” (19.8%). Therefore, compared to providing sufficient MSW facilities, improving residents’ awareness and government services and management are more urgent factors to address in order to encourage the residents to participate in MSWSC.

Figure 3. Concerns with MSWSC.

Another finding was that significant divergences among the eight regions exist regarding satisfaction with the MSW categories (Table 8). The mean percentages of dissatisfaction with the local MSW categories in Beijing, Nanjing, Hangzhou, and Guangzhou were relatively low, especially in terms of the awareness of MSW facilities, at 18.0%, 21.6%, 19.1% and 21.2%, respectively. These cities are all distributed in eastern China, where more financial resources are available in general. Compared with these relatively developed cities, Chongqing, Yichun, Kunming, and Guiyang, which have lower per capita GDP, made limited investments in sanitation, leading to a lower capacity and a poorer quality of the environmental infrastructure in these cities, according to Chen, Geng, and Tsuyoshi (2010). These factors might have influenced residents’ satisfaction with the MSW facilities in these cities (26.0%, 35.7%, 27.33%, and 41.58%, respectively).

Table 8. Comparison of the satisfaction with the MSWSC in different cities.

	A1	A2	A3	B1	B2	B3	C1	C2	C3
Beijing	16.33%	20.41%	17.35%	21.43%	27.55%	65.31%	45.92%	33.67%	17.35%
Nanjing	14.71%	22.55%	27.45%	20.59%	33.33%	60.78%	23.53%	26.47%	28.43%
Hangzhou	11.65%	25.24%	20.39%	11.65%	24.27%	46.60%	28.16%	25.24%	8.74%
Chongqing	16.00%	32.00%	30.00%	16.00%	18.00%	72.00%	51.00%	32.00%	27.00%
Yichun	33.33%	38.38%	35.35%	27.27%	22.22%	67.68%	45.45%	25.25%	18.18%
Guiyang	19.00%	32.00%	31.00%	12.00%	16.00%	78.00%	54.00%	45.00%	34.00%
Kunming	36.63%	44.55%	43.56%	18.81%	24.75%	77.23%	50.50%	34.65%	32.67%
Guangzhou	10.89%	32.67%	18.81%	19.80%	19.80%	63.37%	32.67%	20.79%	16.83%

Notes: A1 is “The irrationality of types of trash can”; A2 is “The inadequacy of the number of trash can”; A3 is “The inaccessibility of trash can”; B1 is “The negative atmosphere of neighbor MSWSC behavior”; B2 is “The ignorance for significance of MSWSC”; B3 is “The lack of concern to MSWSC”; C1 is “The insufficiency of propaganda and education”; C2 is “The backwardness of local MSW services”; C3 is “The imperfection of relevant laws”.

Therefore, on the one hand, the Chinese government should take a positive role in enhancing the public awareness about MSWM improvement via well-designed educational and publicity programs, and on the other hand, it should thoughtfully provide sufficient and convenient separating facilities and supportive infrastructures (Zhang and Wen 2014).

Rewards for MSWSC

As shown in Figure 4, of all the respondents, 65.5% reported they had accepted a reward for MSWSC, while 34.5% claimed that there were no incentives of any kind for MSWSC in their community or city. Respondents in Nanjing gave the highest score (89.2%) on this question, and the lowest three rates were all below10%, which were for Yichun (9.1%), Kunming (6.9%) and Beijing (6.1%). Surprisingly, Beijing, with a relatively high investment rate in MSWSC and a longer duration of implementing the program, had the worst performance in rewarding households for separating MSW. Overall, as the duration of the MSWSC program’s implementation increases, the coverage rate of incentives for MSWSC increases. Additionally, according to Xu et al. (2015) and Zhang et al. (2012), rewarding households for MSWSC can help transform residents’ awareness into behavior; therefore, it is important to provide adequate incentives for MSWSC.

Figure 4. Coverage of rewards for MSWSC.

Performance on MSWSC

Figure 5 shows that in reality, the respondents misclassified more than half of the MSW items (54.0%) rather than separating them into corresponding trashcans and that differences exist among the eight cities. The best performance on MSWSC was found in Nanjing (75.7%) and the three worst performance on MSWSC were in Yichun (20.1%), Kunming (27.4%) and Guiyang (37.1%). The rates in the remaining cities all exceeded average levels (46.0%). These findings related to the progress of MSWSC in the research sites indicated that in practice, the respondents in the earliest pilot cities, Beijing (20 years), Nanjing (16 years), Hangzhou (16 years) and Guangzhou (24 years), separated more MSW into the corresponding trashcans. However, interviewees in Yichun (1 year) and Guiyang (6 years), the two cities with the shortest time since implementing the MSWSC program, were more likely to misclassify MSW. Comparing two cities adopting the MSWSC program between 2000 and 2010, Chongqing had a better performance on MSWSC than did Kunming, while Kunming had a longer time since MSWSC implementation. This result is different from the relationship between the performance on MSWSC and the duration of MSWSC in other cities, and we assume that it could be related to other variables influencing the performance on MSWSC.

Figure 5. Performance on MSWSC.

HLM estimation results

Null model

The null model was built to estimate whether and to what degree MSWSC behavior was influenced by the city-level factors. According to the estimated results of the null model of MSWSC, the fixed and random effects both passed the significance test (p < 0.01). Formula (3) and formula (4) show that the individual and city variables accounted for 82.62% and 17.38% (Table 9) of the variation in the performance on MSWSC, respectively. This result clearly indicates that individual variables are the main factors affecting MSWSC performance. However, it is still necessary to establish the HLM model to evaluate the effect of the variables at each level as the between-group difference (17.38%) exceeds 5.90% (Cohen 1988).

Table 9. Estimated results of the null model of the MSWSC.

	Fixed effect				Random effect
	Parameter	Correlation coefficient	T-ratio	P	Parameter	Standard deviation	Variance component	P	Within-group/between-group difference
Level-1： Y = B0+ R					R	7.308	11.237		82.62%
Level-2： B0 = G00 + U0	G00	8.769	7.728	0.000	U0	3.352	53.403	0.000	17.38%

Random effects regression model

The random effects regression model conducts regression analysis with the first-level variables only. According to the fixed effect results (Table 10), gender (GE, B = 0.1414, P = 0.011), age (AG, B = 0.1145, P = 0.050), accuracy of MSWSC in simulation (AM, B = 0.2280, P = 0.001), satisfaction with sanitation (SS, B = 0.1190, P = 0.050) and satisfaction with publicity (SP, B = 0.1291, P = 0.036) had the highest positive effect on MSWSC behavior. These variables were introduced into the random effect regression model and are discussed further in the next model.

Table 10. Results of the random effects regression model.

Variable	Fixed effect	Coefficient	P-value	Random effect	Chi-square	P-value
GE	G10	0.141	0.010	U10	6.696	>0.500
AG	G20	0.115	0.050	U20	12.653	0.080
EL	G30	−0.028	0.489	U30	2.596	>0.500
FM	G40	−0.007	0.877	U40	9.880	0.195
EN	G50	0.072	0.192	U50	9.562	0.214
KM	G60	0.070	0.156	U60	8.536	0.287
AM	G70	0.228	0.001	U70	14.815	0.038
SS	G80	0.119	0.050	U80	15.852	0.026
SM	G90	0.063	0.212	U90	10.509	0.161
SP	G100	0.129	0.036	U100	14.179	0.048

The results revealed that females outperform males in MSWSC performance, which is consistent with the findings of Meneses and Palacio (2005) and Nixon and Saphores (2009). This is likely due to females’ engagement in household duties (Singhirunnusorn, Donlakorn, and Kaewhanin 2012). Regarding the impact of age on MSWSC behavior, it should be noted that although senior members of the family had good MSWSC performance, their educational attainment and economic income are generally low in China; thus, education level (EL, B = −0.0284) and family monthly income (FM, B = −0.0074) had a slightly negative effect on MSWSC.

According to our results, AM had a significantly positive effect on MSWSC. Thus, to promote MSW recycling programs, it may be necessary to improve residents’ awareness of MSWSC issues (Zhu et al. 2009). Moreover, satisfaction with sanitation and satisfaction with publicity had a significantly positive effect on MSWSC. Evidence in previous literature suggests that the inefficient use of containers and the low popularity of MSWSC may be reasons for citizens’ low levels of satisfaction with the local MSWM (Babaei et al. 2015). However, sanitation and publicity may be different even in a single city or community. Therefore, satisfaction may be affected not only by subjective judgments, but also by objective conditions. In future studies, objective measures of MSWSC should be included in city-level variables.

The random effects of AM (P = 0.038), SS (P = 0.026) and SP (P = 0.048) passed the significance test. The regression coefficients for the three variables had significant variations at the city level. In contrast, the random effects of GE (P > 0.5) and AG (P > 0.5) failed to pass the significance test. These variables are similar in different cities, and the correlation between MSWSC and these variables is not dependent on the city level.

Full model

To create a full HLM and study the coefficient variations of AM, SS, and SP among different cities, per capita gross domestic product (PC), investment in sanitation (IS), reward for MSWSC (RM) and duration of MSWSC (DM) were introduced. Before that, an exploratory analysis was carried out to introduce suitable city-level variables and thus refine the model (Table 11).

Table 11. Results of the exploratory analysis.

	PC	IS	RM	DM
AM	−0.036	0.017	0.021	0.016
SS	0.076	0.009	0.038	0.018
SP	−0.064	−0.003	−0.005	0.006

First, IS, RM and DM had positive effects on the slope of AM. An optimal support management system, including services and funding provided by the city and community, will increase the prevalence of MSWSC training. Residents in the studied cities also tended to have high cognition and judgment abilities. Thus, IS, RM and DM could strengthen the relationship between AM and MSWSC. Although residents in cities with high per capita GDP have a high awareness and knowledge of MSWSC, they spend little time participating in MSWSC and making the optimal choices. Therefore, the correlation between AM and PM was reduced by PC.

Second, the fixed effects show that AM, RM, and DM all played active roles in strengthening the relationship between SS and PM. Communities in developed cities that enjoy investment and reward support have access to better MSW facilities, and residents of these cities are more willing to participate in MSWSC than those of other cities; thus, governmental subsidies have improved residents’ satisfaction with sanitation to some degree. Moreover, a longer duration since the MSWSC program’s implementation means that separation facilities are more suitable for and more accepted by residents. Therefore, DM could strengthen the relationship between SS and PM.

Third, PC had a positive effect on the slope of SP. Economically advanced residents grasp and utilize support policies related to their own interests, so they tend to make higher demands of the local government to improve MSWM.

After adding the above three city-level variables, the random effects failed to pass the significance test. These results showed that the individual-level variables were well explained in the second level and that adding a higher level of variables for further explanation was not necessary. The formulae in the final model are as follows:

Level-1 Model:

(11) $Y=B_0+B_1\left(GE\right)+B_2\left(AG\right)+B_3\left(AM\right)+B_4\left(SS\right)+B_5\left(SP\right)+R$(11)

Level-2 Model:

(12) $B_0=G_{00}+U_0$(12)

(13) $B_1=G_{10}$(13)

(14) $B_2=G_{20}+U_2$(14)

(15) $B_3=G_{30}+G_{31}\left(PC\right)+G_{32}\left(IS\right)+G_{33}\left(RM\right)+G_{34}\left(DM\right)+U_3$(15)

(16) $B_4=G_{40}+G_{41}\left(PC\right)+G_{42}\left(RM\right)+G_{43}\left(DM\right)+U_4$(16)

(17) $B_5=G_{50}+G_{51}\left(PC\right)+U_5$(17)

The influencing factors, including gender, age, knowledge of MSWSC, satisfaction with sanitation and satisfaction with publicity significantly affected participation in MSWSC (at the 5% significance level). All these factors had a positive effect on the dependent variables. In addition, the correlation of PM with the education level and family income was negative. At the city level, critical variables including per capita GDP, investment in sanitation, the degree of rewards for MSWSC and the duration since the MSWSC program’s implementation strengthened or weakened the relationship between MSWSC performance and the individual variables. Finally, although there is a significant association between these influencing factors and MSWSC, it is possible that the relationship is influenced by other variables. Therefore, it would be meaningful to explore other variables can influence MSWSC in future studies, such as home ownership and residential area (Berger 1997), which can influence MSWSC.

Conclusions and policy implications

This study investigated the MSWSC knowledge and behavior of residents of eight pilot cities in China. Ten influencing variables were examined on both the individual and city levels. Based on our results, we present the following conclusions:

1. The individual and city variables accounted for 82.62% and 17.38% of the variation in the performance on MSWSC, respectively. It means that the MSWSC is mainly attributed to individual-level factors, but city-level MSWSC variables should not be ignored.

2. Individual variables, including gender, age, residents’ satisfaction with sanitation and publicity have significantly positive effects on the performance on MSWSC. Specifically, females outperform males in both participation and accuracy in MSWSC practice, and seniors also outperform younger generations.

3. The results of the full model at the city level showed that rewards for MSWSC could strengthen the positive correlation between satisfaction with sanitation and public MSWSC behavior, whereas investment in sanitation could not significantly strengthen this correlation. Moreover, investment in sanitation and rewards for MSWSC could strengthen the positive correlation between the accuracy of MSWSC and public MSWSC behavior.

4. The eight studied cities showed different MSWSC performances. Certain cities showed better accomplishments, such as Nanjing, while certain challenges were also identified in cities such as Yichun and Guiyang because of the worse performance on MSWSC than other pilot cities.

Based on the above conclusions, we present several policy recommendations for decision makers to use in gradually establishing robust MSWSC systems and narrowing nationwide MSWSC service gaps:

1. Females outperform males in both their participation and accuracy in MSWSC practice. This finding highlights that females’ engagement in household duties and the role of females in MSWSC. Senior citizens were also found to outperform younger generations, which is partially due to greater availability and more flexible time. Given that women and senior citizens are the main force for MSWSC in China, specific policies and regulations should be targeted towards other social groups to improve their poor performance on MSWSC.

2. Compared to other barriers to implementing the MSWSC program, such as the poor neighbors’ performance and the unsatisfactory MSWM, residents’ dissatisfactory with MSW disposal facilities was relatively low, indicating that increasing the number and optimizing the distribution of trashcans in public places are not enough.

3. The interest of public is in compensation, so it is crucial to provide economic rewards for MSWSC as a measure by which to encourage residents to learn about MSW classification. Additionally, specialized education and training workshops are required to enhance the awareness of MSWSC. More funding resources should be allocated to more public welfare, such as volunteering and non-governmental organization activities, which can be helpful for residents to learn about MSW classification.

4. Nationwide, efforts should be made to coordinate the development of the recycling industry across cities. Increasing professionalization, obtaining clearer sorting standards and fair allocation of funding resources to regional MSWSC projects and programs can be learned from the experiences of best-performing pilot cities.

Additional information

Funding

This research was supported by “Shu Guang” of the Shanghai Municipal Education Commission and the Shanghai Education Development Foundation (Grant NO. 16SG24).

Notes on contributors

Can Zou

Can Zou is a PhD candidate in the School of Ecological and Environmental Sciences, East China Normal University, Shanghai, China.

Jun Tai

Jun Tai is a professorate senior engineer in the Shanghai Environmental Engineering Design Research Institute, Shanghai, China.

Yu Wang

Yu Wang is a master in the School of Ecological and Environmental Sciences, East China Normal University, Shanghai, China.

Fengyun Sun

Fengyun Sun is a Post Doctorate in the School of Ecological and Environmental Sciences, East China Normal University, Shanghai, China.

Yue Che

Yue Che is a professor in the School of Ecological and Environmental Sciences, East China Normal University, Shanghai, China.

Appendix

English translation of key questions in the questionnaire

Part 1 Resident’s attitudes

1. Which of the following describes the current MSW categories in your city?

   1. Recyclables and other waste

   2. Recyclables, hazardous waste, and other waste

   3. Recyclables, kitchen waste, and other waste

   4. Recyclables, hazardous waste, kitchen waste, and other waste

2. Are there rewards for MSWSC in your residential community?

   1. Yes

   2. No

3. Which of the following has an impact on MSW source separation? (multiple choice)

   1. The irrationality of types of trash can

   2. The inadequacy of the number of trash can

   3. The inaccessibility of trash can

   4. The negative atmosphere of neighbor MSWSC behavior

   5. The ignorance for significance of MSWSC

   6. The lack of concern to MSWSC

   7. The insufficiency of propaganda and education

   8. The backwardness of local MSW education

   9. The imperfection of relevant laws

Part 2 Resident’s behavior

Table

item	classification				actual situation
	A. Recyclable	B. Hazardous waste	C. Kitchen waste	D. Other waste	Sort	Don’t sort
1. Clothing
2. Fruits and vegetables
3. Broken bowl
4. Bad energy-saving lamps
5. Dust
6. Old toys
7. Dry mushroom and jujube
8. Chinese medicine and tea
9. Toilet paper
10. DVD
11. Flour
12. Old shoes
13. Old U disk
14. Expired medicines
15. Paint bucket
16. Baby diapers
17. Chili paste
18. Expired snacks
19. Old newspaper
20. Edible oil drum
21. Medicine bottle
22. Can
23. X-ray
24. Flower
25. Packing box
26. Rechargeable batteries
27. Expired pork

Part 3 Personal attributes

1. 1. Gender

   • A. Male B. Female

2. 2. Age

   • A. 12–17 B.18–25 C.26–35 D.36–45 E.46–55 F. Over 56

3. 3. Education level

   1. Junior high school or lower

   2. Senior high school

   3. Junior high school

   4. Undergraduate

   5. Graduate or above

4. Family monthly income

   1. 3,000 or lower

   2. 3,000–6,000

   3. 6,000–10,000

   4. 10,000–16,000

   5. 16,000 or above

5. Are you or your family members environmentalists?

   1. Oneself

   2. Family members

   3. Neither