*3.4. Sample Characteristics*

The empirical research was conducted in two stages. Firstly, data were collected by a preliminary survey (300 people) in September 2017 as a pretest; secondly, a main survey (2100 people) was conducted through an online platform (Wen Juan Xing). The present research was analyzed based on the data accumulated from the main survey. The main survey, which involved various groups of adults aged between 18 and 70 years, was conducted between December 2017 and the end of January 2018. A sample of 2100 adults between 18 and 69 years old was randomly selected from the local population. After the elimination of responses with outliers and missing values, the number of valid responses was 612 for Shanghai, 484 for Shenyang, and 525 for Chengdu (1621 in total).

The demographic composition of the sample is shown in Table 2. In terms of the respondents' gender, the sample contains slightly more females than males (819 women vs. 802 men). As for the

age category, the number of respondents in their 40s is the largest, and the number of respondents in their 60s is significantly smaller (505 respondents aged 18 to 29 years, 439 respondents in their 30s, 535 respondents in their 40s, 126 respondents in their 50s, and 16 respondents aged 60 years and over). The smaller number could be attributed to the low rate of internet use among seniors. The sample consists of 124 respondents who have achieved a middle school diploma or obtained a lower educational achievement, 456 respondents with a high school or specialized school diploma, 944 respondents with a university or advanced specialized school diploma, and 97 respondents with a graduate school diploma. It is thus clear that 54% of the respondents have completed at least a university-level education. The sample is more or less evenly split between those who have political affiliation and vice versa: 308 Communist Party of China members, 348 Communist Youth League members, 21 members of other non-communist or minor parties, 879 respondents with no political affiliation, and 65 respondents identified as 'other.' The number of respondents who had a household registration for their area of residence (referred to as 'natives') is 916, which is higher than the number of respondents who did not have a household registration for their area of residence (referred to as 'outsiders'), which is 705. The gender ratio, education, and income level are all consistent with the Statistical Yearbook. Hence, the survey reflects the real situation in the three cities under consideration.




**Table 2.** *Cont*.

Source: authors' own calculations, 2020.

### **4. Data Analysis and Results**
