*3.2. Measures*

## 3.2.1. Criterion Variable

The criterion variable in this study was trust in government. In previous studies, researchers have measured overall trust in governmen<sup>t</sup> as a function of participants' trust in various hierarchies of the government, such as the central governmen<sup>t</sup> and local government [5,49]. The 2019 CSS asked participants about their level of trust in central government, county government, and township government. The answers ranged from "no trust at all (1)" to "a grea<sup>t</sup> deal of trust (5)". We used the average value of participants' trust in central, county, and township governments as the level of overall trust in government, with higher scores reflecting greater levels of overall trust in government. Cronbach's α coefficient for overall trust in governmen<sup>t</sup> was 0.744. The KMO value was 0.566 (>0.5), and Bartlett's test was significant (*p* < 0.001), indicating that the three items (trust at each level of government) were suitable for factor analysis [50]. The results of the principal component analysis (PCA) showed that one factor with an eigenvalue greater than 1.0 was retained, and it accounted for 67.244% of the total variance. Additionally, the factor loadings of the three items were 0.620, 0.922, and 0.885, respectively. The results of the confirmatory factor analysis (CFA) showed that the construct reliability (CR) of the three items was 0.784, and the average variance extracted (AVE) was 0.589, indicating that the scale had acceptable convergen<sup>t</sup> validity.

## 3.2.2. Predictor Variable

The predictor variable in this study was social security fairness. Social security is a general term used to refer to various social measures. In this study, social security fairness mainly refers to fairness in terms of public health, employment, and elder security. We measured social security fairness by asking participants the following question: "What do you think of the fairness of the following aspects in current social life: (a) public health, (b) work and employment opportunities, and (c) social security benefits such as elder security?" Respondents answered the question using a five-point rating scale: "very unfair (1)", "generally unfair (2)", "neither unfair nor fair (3)", "generally fair (4)", and "very fair (5)". We used the average scores relating to public health fairness, employment fairness, and elder security fairness to represent the level of social security fairness. Higher scores indicated better social security fairness. Cronbach's α coefficient for social security fairness was 0.659. The KMO value was 0.655 (>0.5), and Bartlett's test was significant (*p* < 0.001), indicating that the three items were suitable for factor analysis. The results of the PCA showed that one factor was extracted which accounted for 59.453% of the total variance. Additionally, the factor loadings of the three items were 0.787, 0.744, and 0.781, respectively. The CFA results indicated that the AVE value was 0.394, and the CR value was 0.667.

#### 3.2.3. Mediator Variables

The first mediator variable in this study was social security satisfaction. Social security satisfaction was measured as overall satisfaction using three social security items: public health security, elder security, and employment security. To measure social security satisfaction, participants were instructed to "Please use a score of 1–10 to express your evaluation

of the following social security items provided by the governmen<sup>t</sup> to the people, where 1 means very dissatisfied and 10 means very satisfied: (a) public health security, (b) elder security, and (c) employment security." In keeping with the 5-point rating scale used above, we converted the 10-point rating scale to a 5-point rating scale, whereby we coded scores of 1 and 2 as "1" and scores of 9 and 10 as "5". A score of "1" meant "very dissatisfied", while a score of "5" meant very satisfied. We took the average satisfaction with the three aspects as the index to measure the level of social security satisfaction. Cronbach's α coefficient for social security satisfaction was 0.837. The KMO value was 0.713 (>0.5), and Bartlett's test was significant (*p* < 0.001). The results of PCA showed that one factor was extracted which accounted for 75.457% of the total variance. The factor loadings of the three items were 0.888, 0.883, and 0.833, respectively. The CFA suggested that the AVE value was 0.387, and the CR value was 0.749.

The second mediator variable in this study was life satisfaction. Life satisfaction was measured as a function of the participants' satisfaction with family relationships, family economic status, education level, leisure, and social life. We converted the 10-point rating scale to a 5-point rating scale, ranging from "very dissatisfied (1)" to "very satisfied (5)". The average level of satisfaction with the five items was used to indicate the level of life satisfaction. Higher scores indicated that participants had greater satisfaction with their lives. Cronbach's α coefficient for life satisfaction was 0.741. The KMO value was 0.756 (>0.5), and Bartlett's test was significant (*p* < 0.001). The results of the PCA indicated that one factor was extracted which accounted for 49.946% of the total variance. The factor loadings of 5 items ranged from 0.491 to 0.811. The CFA suggested that the measurement of life satisfaction had acceptable convergen<sup>t</sup> validity (AVE = 0.637 and CR = 0.843).

#### 3.2.4. Control Variables

We included gender (1 = male and 0 = female), age, education level (1 = senior high school or above and 0 = below senior high school), marital status (1 = married and 0 = not married or divorced), political status (1 = member of the Communist Party of China and 0 = others), region (1 = urban and 0 = rural), Internet use (1 = yes and 0 = no), and location (1 = in the east or west and 0 = others) in the model as control variables.

## *3.3. Statistical Analysis*

We used SPSS 24.0 and Process 2.16 to conduct the statistical analyses. We employed descriptive statistics to examine the overall characteristics of the criterion and predictor variables. Correlation coefficients were computed to examine the strength of linear relationships among social security fairness, social security satisfaction, life satisfaction, and trust in government. Model 6 in Process 2.16 was used to test the serial mediation effects of social security satisfaction and life satisfaction on the relationship between social security fairness and trust in governmen<sup>t</sup> at the central, county, and township levels.
