General characteristics of the participants
Table 1 presents the descriptive statistics of the main sociodemographic characteristics of the study participants used in this study: 68,207 economically active adults aged 19-65 years. The analysis results for gender, education level, region of residence, type of household, marital status and occupation are as follows (in decreasing order of frequency): female (55.3%), male (44.7%); ≥ university (66.4%), high school (27.7%), ≤ high school (5.9%); non-metropolitan areas (66.4%), metropolitan areas (41.6%); co-residential households (88.8%), single-person households (11.2%); married (64.1%), single (25.7%), divorced/widowed/divorced (10.2%); service workers (61.5%), office workers (38.5%).
Correlation analysis and multicollinearity between key variables
Table 2 presents the results of the correlation analysis between the main variables. The correlation coefficients between the main variables did not exceed the threshold of 0.8, with absolute values ranging from 001 to 0.637. Similarly, with the variance inflation factors values ranging from 1,040 to 2,203, well below the threshold of 10, there was no problem of multicollinearity.
Mediating and dual mediating effects of COVID-19 related changes in daily life changes and anxiety
For two mediator variables, as is the case with this research model, a three-step regression analysis is performed using Model 6 of PROCESS macro to analyze the simple and double mediating effects. Accordingly, we analyzed the dual mediation of variations in daily life due to COVID-19 and anxiety induced thereby in the mental health relationship between occupation and depression, depending on the type of occupation. The analysis results are shown in Table 3.
Step 1 of Table 3 shows the results of multiple linear regression analysis regarding the effect of the independent variable on mediator variable 1 (changes in daily life due to COVID-19), which was performed as the first process step of data analysis in Model 6 The explanatory power of the model was 11.2%, which was statistically significant (F = 93.247, p < .001). In this model, the independent variable – occupational type – was found to have a negative effect on mediator variable 1 – changes in daily life due to COVID-19 (B = − 2,199, p < .001). It meant that service workers experienced a higher level of perceived changes in daily life as a result of COVID-19 than office workers. For the control variables, the following groups reported more acutely perceived changes in daily life due to COVID-19 than the opposite groups: women (B = -3,198, p < .000), younger age groups (B = -.134, p < .000), metropolitan area (B = 1.041, p<. 000), single-person households (B = 1,172, p < .000), and a lower income (B = -.930, p < .000).
Step 2 of Table 3 shows the results of analyzes regarding the effects of the independent variable (occupational type) and mediator variable 1 (changes in daily life due to COVID-19) on mediator variable 2 (anxiety caused by COVID-19), which were performed as the second process step of data analysis in Model 6. The step 2 analysis model was also statistically significant (F = 569,948, p < .001), where the independent variable had a negative effect on moderator variable 1 (B = -.043, p < .001), and moderator variable 1 had a positive effect on moderator variable 2 (B = .006, p < .001), both with statistical significance. This shows that the fear due to COVID-19 increased to a greater extent among service workers than among office workers and that the greater the degree of perceiving changes in daily life due to COVID-19, the greater the fear caused by COVID-19 -19. As for the control variables, the following groups reported a greater increase in anxiety caused by COVID-19 compared to the opposite groups: women (B = -.180, p < .000), older age groups (B = .001, p < .000), at lower education level (B = -.098, p < .000), non-urban areas (B = -.076, p < .000), co-residential households (B = -.057, p < .000), and a lower income (B = -.033, p < .000).
Finally, step 3 of Table 3 shows the results of analyzes regarding the effects of the independent variable and mediator variables 1 and 2 on the dependent variable, depression, which were performed as the third step of data analysis in model 6. Similarly, step 3 analysis model was statistically significant (F = 233,661, p < .001), which found that depression was significantly influenced by occupational type (B = -.011, p < .001), changes in daily life due to COVID-19 (B = .001, p < .001), and anxiety due to COVID-19 (B = .017, p < .001). In other words, depression increased more in service workers than in office workers. The greater the degree of perceiving changes in daily life due to COVID-19 and the higher the level of anxiety caused by COVID-19, the greater the increase in depression due to COVID-19. Furthermore, regarding the control variables, the following groups reported a greater increase in depression compared to the opposite groups: women (B = -.044, p < .000), younger age groups (B = -.003, p < .000), lower educational level (B = -.013, p < .000), metropolitan area (B = .030, p < .000), single-person households (B = .054, p < .000), single (B = .018, p < .000), and a lower income (B = -.017, p < .000).
Figure 2 is a schematic summary of the analysis results for the dual mediating effects of COVID-19-related changes in daily life and anxiety on the relationship between occupational type and depression, as shown in Table 3.
Table 4 presents the results of statistical significance tests of mediating effects via the bootstrap method performed after completion of the 3-step regression analysis performed in Model 6 of PROCESS macro. The results confirm the statistical significance of the overall mediating effect (.0032; Boot CI: .0038 to −.0026), establishing the model fit of the dual mediation model evaluated in this study. Regarding the mediating effects of each pathway, the effect of occupational type on depression via mediator variable 1 – changes in daily life due to COVID-19 – was found to be statistically significant (Boot CI: −.0027 to -.0018), since the effect was via mediator variable 2 – anxiety due to COVID-19 (Boot CI: .0010 to .0005). Similarly, the dual mediating effects, which the occupational type has on depression from changes in daily life and anxiety due to COVID-19, were found to be statistically significant (−.0002; Boot CI: .0003 to .0002).