**4. Results**

#### *4.1. Descriptive Statistics and Correlation Analysis*

Table 1 shows descriptive statistics and correlation coefficients. The average score for Chinese citizens' overall trust in their governmen<sup>t</sup> was 3.910. The central governmen<sup>t</sup> enjoyed a higher level of trust than the county and township governments (M = 4.492, M = 3.745, and M = 3.494, respectively). Paired sample *t*-tests showed the three means differed significantly: (a) the mean difference between trust in central and county governments was 0.744 (*t* = 56.975, *df* = 7402, *p* < 0.001, medium Cohen's *d* = 0.662), (b) the mean difference between trust in central and township governments was 0.998 (*t* = 65.466, *df* = 7402, *p* < 0.001, medium Cohen's *d* = 0.761), and (c) the mean difference between trust in county and township governments was 0.251 (*t* = 25.905, *df* = 7402, *p* < 0.001, small Cohen's *d* = 0.301). The average score of social security fairness across all levels was 3.491. Scores concerning citizens' satisfaction with social security and life were also at a similar level (M = 3.453 and M = 3.471, respectively).


**Table 1.** Descriptive statistics and correlations among the variables.

\*\*\* *p* < 0.001.

The correlation analysis showed that social security fairness was positively associated with overall trust in governmen<sup>t</sup> (*r* = 0.401, *p* < 0.001). Social security fairness was significantly associated with trust in central, county, and township governments (*r* = 0.189, *p* < 0.001; *r* = 0.375, *p* < 0.001 and *r* = 0.387, *p* < 0.001, respectively). Social security satisfaction and life satisfaction were significantly positively associated with overall trust in governmen<sup>t</sup> (*r* = 0.375, *p* < 0.001 and *r* = 0.259, *p* < 0.001, respectively). Social security satisfaction was also significantly positively associated with life satisfaction (*r* = 0.441, *p* < 0.001). Correlations among social security fairness, social security satisfaction, life satisfaction, and trust in governmen<sup>t</sup> were all significant. We also found that social security fairness, social security satisfaction, and life satisfaction had the weakest correlations with trust in central governmen<sup>t</sup> and the strongest correlations with trust in township government. The correlations between trust and other variables were higher for lower levels of government. Considering that correlations were significant among the variables, we performed several mediation analyses.

#### *4.2. The Serial Mediation Effects of Social Security Satisfaction and Life Satisfaction*

We used Amos software to analyze the overall fit of the tested models before path analysis. The results presented acceptable model fit indices (CFI = 0.946, TLI = 0.931, RMSEA = 0.060, SRMA = 0.034, and chi-square/*df* = 27.8). We used the bootstrap sampling method to test the serial mediation effect through Model 6 in the Process 2.16 plug-in of the SPSS macro program. The sample size was set to 5000, and the confidence level was 95%. Mediation analyses included the following control variables: gender, age, education level, marital status, political status, region, Internet use, and location. Figure 1 shows the results of the path analysis. The proposed model explained 23.8% of the variance in social security satisfaction (*p* < 0.001), 27.1% of the variance in life satisfaction (*p* < 0.001), and 22.7% of the variance in overall trust in governmen<sup>t</sup> (*p* < 0.001). The results demonstrated that social security fairness had a positive and statistically significant direct effect on overall trust in governmen<sup>t</sup> (β = 0.286, *p* < 0.001). The path coefficient between social security satisfaction and social security fairness was 0.472 (*p* < 0.001), indicating that social security fairness significantly positively predicted social security satisfaction. The path coefficient between overall trust in governmen<sup>t</sup> and social security satisfaction was 0.192 (*p* < 0.001), showing that social security satisfaction significantly partially mediated the relationship between social security fairness and overall trust in governmen<sup>t</sup> (β = 0.091, *p* < 0.001). In addition, the 95% confidence intervals of bootstrapping with a sample size of 5000 were 0.077 and 0.104, excluding 0.

The path coefficient between social security fairness and life satisfaction was 0.078 (*p* < 0.001), indicating that social security fairness significantly positively predicted life satisfaction. The path coefficient between life satisfaction and overall trust in governmen<sup>t</sup> was 0.079 (*p* < 0.001). Life satisfaction partially mediated the association between social security fairness and overall trust in governmen<sup>t</sup> (β = 0.006, *p* < 0.001). In addition, the 95% confidence intervals of bootstrapping with a sample size of 5000 were 0.004 and 0.010, excluding 0.

**Figure 1.** The serial mediator model of social security fairness, social security satisfaction, life satisfaction, and overall trust in governmen<sup>t</sup> after adding the control variables (*n* = 7403). Standardized regression coefficients are shown next to the arrows. Adjusted R<sup>2</sup> is shown above the explained variable. \*\*\* *p* < 0.001.

The path coefficient between social security satisfaction and life satisfaction was 0.385 (*p* < 0.001), indicating that life satisfaction was highly correlated with social security satisfaction. The results revealed that the serial mediation effects of social security satisfaction and life satisfaction between social security fairness and overall trust in governmen<sup>t</sup> were significant (β = 0.014, *p* < 0.001). The 95% confidence intervals of bootstrapping with a sample size of 5000 were 0.010 and 0.019, excluding 0. Therefore, all path coefficients in the model reached the level of statistical significance (*p* < 0.001). Social security fairness indirectly partially predicted overall trust in governmen<sup>t</sup> through social security satisfaction, life satisfaction, and the serial mediation of social security satisfaction and life satisfaction.

We further examined the serial mediation effects that social security satisfaction and life satisfaction had on the relationship between social security fairness and trust in governmen<sup>t</sup> at the central, county, and township levels. Figure 2 presents the results of the path analysis between social security fairness and trust in central government. After adding control variables, the path coefficient between social security fairness and trust in central governmen<sup>t</sup> was 0.134 (*p* < 0.001), indicating that social security fairness directly and positively predicted trust in central government. The path coefficient between social security satisfaction and trust in central governmen<sup>t</sup> was 0.113 (*p* < 0.001), showing that social security satisfaction partially mediated the relationship between social security fairness and trust in central governmen<sup>t</sup> (β = 0.054, 95% CIs: 0.039, 0.067). Meanwhile, the path coefficient between life satisfaction and trust in central governmen<sup>t</sup> was 0.016 (*p* > 0.05), indicating that the prediction of trust in central governmen<sup>t</sup> using life satisfaction was not significant. Life satisfaction was not a significant mediator in the relationship between social security fairness and trust in central government. The results reveal that social security fairness cannot significantly and indirectly predict trust in central governmen<sup>t</sup> through life satisfaction (95% CIs: −0.001, 0.035) and the serial mediation of social security satisfaction and life satisfaction (95% CIs: −0.002, 0.008). In addition, the serial model explained the change in trust in central governmen<sup>t</sup> by 9.9% (*p* < 0.001).

Figure 3 shows the results of the path analysis between social security fairness and trust in local governmen<sup>t</sup> at the county and township levels. The serial mediation model explained 19.6% of the variance in trust in county governmen<sup>t</sup> (*p* < 0.001). The path coefficients of social security fairness, social security satisfaction, and life satisfaction on trust in county governmen<sup>t</sup> were 0.267 (*p* < 0.001), 0.170 (*p* < 0.001), and 0.083 (*p* < 0.001), respectively. Social security fairness indirectly predicted trust in county governmen<sup>t</sup> through social security satisfaction, life satisfaction, and their serial mediation were 0.080 (95% CIs: 0.067, 0.094), 0.006 (95% CIs: 0.004, 0.010), and 0.015 (95% CIs: 0.010, 0.020), respectively.

**Figure 2.** The serial mediator model of social security fairness, social security satisfaction, life satisfaction, and trust in central governmen<sup>t</sup> after adding the control variables (*n* = 7403). Standardized regression coefficients are marked next to the arrows. Adjusted R<sup>2</sup> is marked above the explained variable. \*\*\* *p* < 0.001.

**Figure 3.** The serial mediator model of social security fairness, social security satisfaction, life satisfaction, and trust in local (county and township) governmen<sup>t</sup> after adding the control variables (*n* = 7403). Standardized regression coefficients are shown next to the arrows. Adjusted R<sup>2</sup> is shown above the explained variable. \*\*\* *p* < 0.001.

The results of the path analysis between social security fairness and trust in township governmen<sup>t</sup> show that the serial mediation model explained 20.7% of the variance in trust in township governmen<sup>t</sup> (*p* < 0.001). The path coefficients from social security fairness, social security satisfaction, and life satisfaction to trust in township governmen<sup>t</sup> were 0.278 (*p* < 0.001), 0.179 (*p* < 0.001), and 0.082 (*p* < 0.001), respectively. Social security fairness significantly and directly predicted trust in township governmen<sup>t</sup> (β = 0.278, *p* < 0.001). Social security fairness and trust in township governmen<sup>t</sup> were related through social security satisfaction (β = 0.085, 95% CIs: 0.072, 0.099), life satisfaction (β = 0.006, 95% CIs: 0.004, 0.010), and their serial mediation (β = 0.015, 95% CIs: 0.010, 0.020), respectively.

In addition, the regression results concerning the control variables revealed some demographic factors that predicted overall trust in government. Since a large sample size can influence the statistical significance of results, *p* = 0.001 was used to evaluate significance. Citizens' age (β = 0.065, *p* < 0.001), education level (β = 0.075, *p* < 0.001), and political status (β = 0.062, *p* < 0.001) were significantly associated with their overall trust in government. The path coefficients from gender (β = −0.008, *p* > 0.001), marital status

(β = −0.012, *p* > 0.001), region (β = −0.03, *p* > 0.001), Internet use (β = −0.03, *p* > 0.001), living in eastern China (β = 0.023, *p* > 0.001), or living in western China (β = −0.007, *p* > 0.001) to overall trust in governmen<sup>t</sup> were not significant at the 0.001 level. Citizens who are older, have a higher education level, and are members of the Communist Party of China have a higher level of overall trust in government. We can also see that social security fairness is capable of significantly and positively predicting trust in central governmen<sup>t</sup> (β = 0.134, *p* < 0.001), trust in county governmen<sup>t</sup> (β = 0.267, *p* < 0.001), and trust in township governmen<sup>t</sup> (β = 0.278, *p* < 0.001). In examining the adjusted R<sup>2</sup> changes, the serial mediation model appeared to have a higher explanatory power to trust in county (adjusted R<sup>2</sup> = 0.196, *p* < 0.001) and township (adjusted R<sup>2</sup> = 0.207, *p* < 0.001) governments than in central governmen<sup>t</sup> (adjusted R<sup>2</sup> = 0.099, *p* < 0.001). The results illustrate the fact that the positive prediction of trust in governmen<sup>t</sup> via social security fairness was better for lower levels of the governmen<sup>t</sup> than for higher levels.
