Descriptive statistics and correlation analyses
Table 1 lays out the variables’ baseline sample characteristics and their means and standard deviations over time. The mean age of the sample was 67.51. Women accounted for 45.57%, Han Chinese accounted for 92.39%, and urban residents accounted for 52.16% of the total sample. Other demographics included 97.46% married, 4.93 years (average) of education, and 15.38% members of the CPC. Their ISEI showed a decreasing trend over time, and their AHI, SSS, SRH, and ST levels showed an increasing trend.
Table 1 Sample characteristics at baseline and means and standard deviations of variables over time
Variables
|
Time 1
|
Time 2
|
Time 3
|
Time 4
|
M (%)
|
SD (Range)
|
M
|
SD
|
M
|
SD
|
M
|
SD
|
Age
|
67.51
|
5.77
|
|
|
|
|
|
|
Female
|
(45.57)
|
(0–1)
|
|
|
|
|
|
|
Han Chinese
|
(92.39)
|
(0–1)
|
|
|
|
|
|
|
Non-Agricultural
|
(52.16)
|
(0–1)
|
|
|
|
|
|
|
Married
|
(97.46)
|
(0–1)
|
|
|
|
|
|
|
Education
|
(4.93)
|
4.63
|
|
|
|
|
|
|
CPC membership
|
(15.38)
|
(0–1)
|
|
|
|
|
|
|
ISEI
|
26.18
|
9.29
|
26.70
|
12.70
|
25.97
|
11.73
|
26.00
|
11..73
|
AHI (Unit: RMB)
|
8279.91
|
6,996.36
|
7578.67
|
10605.33
|
8661.27
|
17982.85
|
9435.20
|
11244.71
|
SSS
|
3.25
|
1.26
|
3.04
|
1.24
|
3.23
|
1.28
|
3.48
|
1.24
|
SRH
|
2.68
|
1.33
|
2.52
|
1.35
|
2.66
|
1.33
|
2.87
|
1.29
|
ST
|
3.51
|
1.93
|
3.12
|
2.00
|
3.49
|
1.94
|
3.93
|
1.77
|
n = 4536. CPC = Communist Party of China; ISEI = International Socioeconomic Index; AHI = annual household income; SSS = subjective social status; SRH = self-rated health; and ST = social trust.
Table 2 displays each variable’s mean, standard deviation, and the correlation among key variables in the three measurements (Time 1, 2, and 3). Significant positive correlation existed between SSS, ST, and SRH one another; the coefficients between the variables were all less than 0.8; and the data did not suffer from significant multicollinearity, meeting the prerequisites for the mediation test [34].
Table 2 Correlation analyses between key variables
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
8
|
9
|
T1 SSS
|
1
|
|
|
|
|
|
|
|
|
T2 SSS
|
0.56**
|
1
|
|
|
|
|
|
|
|
T3 SSS
|
0.39**
|
0.38**
|
1
|
|
|
|
|
|
|
T1 ST
|
0.35**
|
0.27**
|
0.20**
|
1
|
|
|
|
|
|
T2 ST
|
0.23**
|
0.23**
|
0.18**
|
0.45**
|
1
|
|
|
|
|
T3 ST
|
0.12**
|
0.12**
|
0.18**
|
0.40**
|
0.50**
|
1
|
|
|
|
T1 SRH
|
0.20**
|
0.12**
|
0.11**
|
0.13**
|
0.15**
|
0.16**
|
1
|
|
|
T2 SRH
|
0.14**
|
0.13**
|
0.13**
|
0.14**
|
0.15**
|
0.16**
|
0.54**
|
1
|
|
T3 SRH
|
0.15**
|
0.15**
|
0.14**
|
0.15**
|
0.17**
|
0.18**
|
0.53**
|
0.56**
|
1
|
Mean
|
3.04
|
3.23
|
3.48
|
2.52
|
2.66
|
2.87
|
3.12
|
3.49
|
3.93
|
SD
|
1.24
|
1.28
|
1.24
|
1.35
|
1.33
|
1.29
|
2.00
|
1.94
|
1.77
|
**p < 0.01. SSS = subjective social status, ST = social trust, SRH = self-rated health, SD = standard deviation.
Trajectories of SSS, ST, and SRH
We used unconditional LGMs to explore the trajectories of the studied variables and individually fit the SSS, ST, and SRH scores. Table 3 describes the fit indicators of the three models and the means of the intercepts and slopes. The study found that the models of SSS, ST, and SRH were well fitted, and the mean slope was positive, indicating that the development of the SSS, ST, and SRH all showed an increasing linear trend. However, the partial correlation coefficients were negative, indicating that the higher the initial level of the SSS, ST, and SRH, the slower the growth rate [35].
Table 3 LGM fitting indicators, intercepts, and slopes of key variables
Variables
|
χ²
|
df
|
RMSEA
|
CFI
|
TLI
|
Mean
|
Partial Correlation
|
Intercept
|
Slope
|
SSS
|
4.762
|
1
|
0.029
|
0.999
|
0.996
|
2.809***
|
0.565***
|
−0.692***
|
ST
|
5.079
|
1
|
0.030
|
0.999
|
0.997
|
2.476***
|
0.789***
|
−0.256***
|
SRH
|
2.486
|
1
|
0.018
|
0.999
|
0.998
|
2.171***
|
0.745***
|
−0.398***
|
*** p < 0.001. SSS = subjective social status, ST = social trust, SRH = self-rated health.
Direct effect of SSS on the initial level and growth rate of SRH
We constructed a conditional LGM, with SSS as the independent variable and SRH as the dependent variable, to examine whether SSS can affect the initial level and growth rate of SRH. Figure 1 shows abbreviated results of the path analysis in this model as the standardized path coefficients of the key variables. The conditional model had a good fit (χ²/df = 183.760/44, CFI = 0.978, TLI = 0.965, RMSEA = 0.026). The SSS intercept significantly and positively influenced the SRH intercept (β = 0.554, p < 0.001) and slope (β = 0.164, p < 0.05), indicating that the higher the initial level of SSS, the higher the initial level of the SRH, and the faster the growth rate of the SRH. The SSS slope significantly and positively affected the SRH slope (β = 0.555, p < 0.001), indicating that the faster the improvement of SSS among the elderly, the faster the improvement of their health.
[Insert Figure 1 about here]
Longitudinal mediation of ST between SSS and SRH
Model establishment and fitting
We used the LGMs of SSS, ST, and SRH to examine the longitudinal mediating role of ST. We tested the mediation of ST by establishing a structural equation model whose fit indicators were also tested. The results revealed that the model had a good fit (χ²/df = 374.085/85, CFI = 0.937, TLI = 0.913, RMSEA = 0.042). Figure 2 illustrates the final model.
[Insert Figure 2 about here]
Direct effects between the variables in the longitudinal mediation model
Table 4 describes the direct paths between the SSS, ST, and SRH. When we controlled the covariates, including gender, age, ethnicity, Hukou, AHI, and ISEI, the intercept of SSS had a significant positive effect on the intercept of ST (β = 0.292, p < 0.001) and the slope of ST (β = 0.577, p < 0.01). This result indicates that the initial level of SSS significantly and positively affected the initial ST level and promoted the ST growth rate. In addition, the slope of SSS significantly and positively impacted the slope of ST (β = 1.259, p < 0.01), indicating that the growth rate of SSS positively contributed to the growth rate of ST. Furthermore, the intercept of SSS had a significant positive effect on the intercept of SRH (β = 0.531, p < 0.001). It also significantly and positively predicated the SRH slope (β = 1.245, p < 0.01), signifying that the higher the SSS of the elderly, the higher the initial level and the faster the growth rate of their health. The intercept of ST also significantly and positively predicated the intercept and slope of SRH (β = 0.091, p < 0.001;β = 0.158, p < 0.001), indicating that the initial level of ST had a significant and positive effect on the initial level and growth rate of health. Finally, the SSS and ST slopes significantly and positively affected the SRH slope, indicating the SSS and ST growth rates promoted the SRH growth rate.
Table 4 Direct paths in the longitudinal mediation model
Path coefficient
|
β
|
B
|
S.E.
|
C.R.
|
p
|
SSS intercept
|
→
|
ST intercept
|
0.292
|
0.268
|
0.020
|
13.642
|
0.000
|
SSS intercept
|
→
|
ST slope
|
0.577
|
0.060
|
0.017
|
3.450
|
0.001
|
SSS slope
|
→
|
ST slope
|
1.259
|
0.391
|
0.076
|
5.126
|
0.000
|
SSS intercept
|
→
|
SRH intercept
|
0.531
|
0.699
|
0.034
|
20.558
|
0.000
|
ST intercept
|
→
|
SRH intercept
|
0.091
|
0.130
|
0.035
|
3.685
|
0.000
|
SSS intercept
|
→
|
SRH slope
|
1.245
|
0.614
|
0.233
|
2.633
|
0.008
|
SSS slope
|
→
|
SRH slope
|
3.706
|
5.442
|
1.550
|
3.512
|
0.000
|
ST intercept
|
→
|
SRH slope
|
0.158
|
0.085
|
0.021
|
4.116
|
0.000
|
ST slope
|
→
|
SRH slope
|
2.499
|
11.804
|
4.045
|
2.918
|
0.004
|
B = unstandardized coefficient, β = standardized coefficient, SE = standard error, CR = critical ratio, SSS = subjective social status, ST = social trust, SRH = self-rated health
Test of the longitudinal mediation effect of ST
We used bootstrapping (1,000 re-samples) to verify the mediating effect of the intercept and slope of ST, respectively, to explore the longitudinal effect of ST in the association between SSS and SRH. The model included four indirect paths: (1) SSS intercept → ST intercept → SRH intercept, (2) SSS intercept → ST intercept → SRH slope, (3) SSS intercept → ST slope → SRH slope, and (4) SSS slope → ST slope → SRH slope. The bootstrapping results showed that the four indirect paths were significant (see Table 5), indicating that SSS indirectly affected the health development of the elderly. In addition, ST’s initial level and growth rate played a longitudinal mediating role between SSS and health. Among them, the ST intercept partially mediated the interplay between SSS and SRH intercepts, accounting for 4.84% of the total effect; the ST intercept partially mediated the linkage between the SSS intercept and SRH slope, accounting for 1.68% of the total effect; the ST slope partially mediated the interaction between the SSS intercept and SRH slope, accounting for 52.83% of the total effect; and the ST slope partially mediated the connection between the SSS and SRH slopes, accounting for 45.91% of the total effect.
Table 5 Indirect paths in the longitudinal mediation model
Indirect Paths
|
Indirect Effect
|
p
|
95% Bootstrap
|
Proportion of Mediating Effect
|
SSS intercept → ST intercept→ SRH intercept
|
0.027
|
0.000
|
0.015
|
0.039
|
4.84%
|
SSS intercept → ST intercept→ SRH slope
|
0.046
|
0.000
|
0..020
|
0.065
|
1.68%
|
SSS intercept → ST slope→ SRH slope
|
1.443
|
0.001
|
1.173
|
1.956
|
52.83%
|
SSS slope→ ST slope → SRH slope
|
3.145
|
0.001
|
1.844
|
4.911
|
45.91%
|
SSS = subjective social status, ST = social trust, SRH = self-rated health.