Raw data of waste sorting behavior scale

1.Questionnaire designTo ensure accuracy and scientific validity, the measurement items used in this study were based on previous literature, with minor modifications made to fit the current research context. The questionnaire underwent four rounds of revision. 1) we translated the questionnaire from English to Chinese and then invited 2 local Chinese individuals with strong English communication skills to modify the language to make it suitable for the target audience. Due to the rigid translation of “I am willing to contribute money to garbage classification”, it was suggested to change the phrase to “I am willing to donate to public welfare activities for garbage classification”. 2) We invited 10 relevant professionals to evaluate the logic and rigor of the initial wording of the questionnaire, which was beneficial for aligning the questionnaire questions with the core concepts. The statement “The government's policies have prompted me to implement waste sorting” was found to be not accurate enough to highlight the meaning of structural policies; thus it was changed to “The government has issued policies related to the nationwide garbage classification campaign, prompting me to engage in garbage classification behavior”. 3) We invited 15 ordinary residents to provide communication and feedback to make the questionnaire more understandable and practical. Due to the resident’s feedback that the term “Social media” was too difficult to understand, we changed the tern to the phrase “Social media (WeChat, Weibo, Tiktok, etc.)”. 4) To test the viability of the initial scale, a pre-survey was conducted between 1 April 2024 and 15 April 2024; approximately 518 valid questionnaires were returned. The analysis of the data showed that the factor loadings of two items were below 0.6, resulting in insufficient reliability. The two items were “Personal norms (PN2): Not sorting garbage would violate my moral principles”, and “Compliance waste sorting behavior (CWSB5): I practice waste sorting due to external pressure”, respectively. In addition, an item is “Social norm (SN2): my garbage classification behavior is affected by the mass media (television, the Internet, etc.)”, and its AVE square root value is 0.768, less than the maximum value (0.773) of the absolute value of the correlation coefficient between factors, which means that its discrimination validity is not good. After the deletion of three items, the loadings of all the variable factors were higher than 0.6 (Bagozzi & Yi, 1988). The Cronbach’s alpha values ranged from 0.808 to 0.924, indicating good reliability. The component reliability (CR) ranged from 0.812 to 0.926 (exceeding the threshold of 0.7) and the average variance extracted (AVE) ranged from 0.559 to 0.743 (exceeding the threshold of 0.5), indicating that each variable has good convergent validity (Larcker, 1981). Furthermore, the square root of AVE is greater than the latent variable’s correlation with another variable in the model, indicating discriminant validity (Fornell & Larcker, 1981). We adjusted the initial order of items in the scale design to prevent identical responses to the same question and inserted lie-detector questions to help screen out invalid questionnaires later. Each variable is measured by 3–7 items scored on a 7-point Likert scale ranging from “Strongly Disagree” to “Strongly Agree” (1 = Strongly Disagree; 7 = Strongly Agree). 2.Data collectionThe questionnaires were collected through a stratified random sampling procedure, which it was divided into the following four steps. Step 1. Determine the stratification criteria. Since 2019, the waste sorting policy has covered 31 provinces in Chinese Mainland. Considering China’s vast territory and diverse demographic characteristics, in order to ensure the representativeness and universality of sampling, we divided mainland China (with the exception of Taiwan Province, Hong Kong and Macao) into four major regions, namely, east, central, west and northeast, based on the geographical location and socioeconomic development of the different regions. Step 2. Calculate the sample size. According to the data from the Seventh National Census, we selected samples based on a population ratio in the east, central, west and northeast regions. Since this study involved random sampling, we used the equation below to calculate the minimum sample size (given a confidence level of 95%, a standard deviation value of 0.5, and an error margin of 5%) to ensure the accuracy of the measurements and results: n=Z2σ2 /E2n: sample size for the study.Z: confidence level. The Z score is 1.96 at the 95% confidence level.Σ: standard deviation.E: margin of error.Formula calculations showed that the minimum sample size needed to draw meaningful conclusions was 384 participants. Step 3. Collecting Online Data. From May to July 2024, formal surveys were conducted via online platforms. Compared to traditional paper-based questionnaires, online surveys can reach a broader pool of potential respondents as they are not constrained by geographical location, thereby collecting more diverse sample data. This not only enhances the external validity of the research but also makes the results more representative. We selected the Sojump online data collection platform (http://www.sojump.com/), which has processed hundreds of millions of online surveys. Its extensive experience and advanced features—such as time controls and automatic invalid response elimination—all contribute to ensuring questionnaire quality.Sojump was used to generate survey URLs, which were randomly distributed via the WeChat China network platform. Strict measures were implemented during data collection to guarantee sample authenticity, including setting IP address restrictions and real-name verification. Only individuals passing these checks could complete the questionnaire. In the informed consent form presented before the survey began, we committed that the survey content would be limited to academic research and that we would not disclose respondents’ personal information. All participants received compensation upon completing the questionnaire.Step 4. Data consolidation. Through mutual consultation among research team members, we identified 1,767 valid questionnaires.