3  Results
We first compared the data collected before versus during the coronavirus outbreak. We found that the outbreak significantly degraded emotional well-being (M before = .437, SD before = .568; M after = .114, SD after = .626; F(1, 14129) = 728.808, p < .001)—a 74% decline. We ran a series of regressions with emotional well-being as the dependent variable, and the coronavirus outbreak (1 = during, 0 = before), whether the individual resided in Hubei (1 = yes, 0 = no), age, sex (1 = female, 0 = male), marital status (1 = married, 0 = not married), household income, and each of their interaction term with the outbreak as predictors (see Table 1 ). The analyses not only established a consistent, significant negative effect of the outbreak on emotional well-being, but also revealed a set of significant interactions: (i) Individuals residing in Hubei, the epicenter of the outbreak, experienced a larger decline in emotional well-being. Because the overwhelming majority of Chinese coronavirus patients resided in that region (Dong, Du, and Gardner, 2020), this result suggests that a higher likelihood of contracting the disease accentuates the detrimental effect of an epidemic outbreak on emotional well-being. (ii) Those of an older age also experienced a larger reduction of emotional well-being during the outbreak. Because the coronavirus tends to cause more harm to the elderly than people of a younger age (CDC, 2020), this pattern suggests that the extent to which an individual might suffer from contracting the disease moderates the effect of an epidemic on the person's emotional well-being. (iii) Individuals who were married also experienced a greater decline in emotional well-being, suggesting that enduring an outbreak (e.g., being in a confined space for extended periods of lockdown) can potentially exacerbate relational issues that worsen emotional well-being. This pattern is consistent with the increase in marriage problems after the COVID-19 outbreak in China (Financial Times, 2020). Neither income or gender had a significant interaction effect with the outbreak.
Table 1 The impact of coronavirus outbreak on emotional well-being
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Model 11 Model 12 Model 13 Model 14 Model 15 Model 16 Model 17
Outbreak -.323⁎⁎⁎⁎ (.012) -.322⁎⁎⁎⁎ (.012) -.316⁎⁎⁎⁎ (.012) -.317⁎⁎⁎⁎ (.012) -.322⁎⁎⁎⁎ (.012) -.318⁎⁎⁎⁎ (.012) -.315⁎⁎⁎⁎ (.012) -.311⁎⁎⁎⁎ (.012) -.149⁎⁎⁎⁎ (.044) -.145⁎⁎⁎⁎ (.044) -.327⁎⁎⁎⁎ (.017) -.323⁎⁎⁎⁎ (.017) -.241⁎⁎⁎⁎ (.021) -.241⁎⁎⁎⁎ (.021) -.301⁎⁎⁎⁎ (.022) -.294⁎⁎⁎⁎ (.022) -.146⁎⁎⁎ (.048)
Hubei .023 (.023) .024 (.023) .024 (.023) .023 (.023) .026 (.023) .054* (.025) .057* (.025) .026 (.023) .026 (.023) .024 (.023) .025 (.023) .056* (.025)
Age .002⁎⁎⁎⁎ (.0004) .002⁎⁎⁎⁎ (.0004) .001 (.0005) .001 (.0005) .001 (.0005) .002⁎⁎⁎⁎ (.0004) .001⁎⁎ (.001) .001 (.005) .0004 (.0005) .001 (.0005) .001 (.001)
Sex .008 (.01) .006 (.01) .006 (.01) .006 (.01) .005 (.01) .005 (.011) .004 (.011) .005 (.01) .006 (.01) .003 (.011)
Married .052⁎⁎⁎⁎ (.013) .04⁎⁎⁎ (.013) .039⁎⁎⁎ (.013) .041⁎⁎⁎ (.013) .04⁎⁎⁎ (.013) .089⁎⁎⁎⁎ (.012) .07⁎⁎⁎⁎ (.015) .041⁎⁎⁎ (.014) .061⁎⁎⁎⁎ (.016)
Income .002⁎⁎⁎⁎ (.0003) .002⁎⁎⁎⁎ (.0003) .002⁎⁎⁎⁎ (.0003) .002⁎⁎⁎⁎ (.0003) .002⁎⁎⁎⁎ (.0003) .002⁎⁎⁎⁎ (.0004) .002⁎⁎⁎⁎ (.0004) .002⁎⁎⁎⁎ (.0004)
Outbreak × Hubei -.183⁎⁎⁎ (.06) -.179⁎⁎⁎ (.06) -.183⁎⁎⁎ (.06)
Outbreak × Age -.005⁎⁎⁎⁎ (.001) -.005⁎⁎⁎⁎ (.001) -.003* (.001)
Outbreak × Sex .008 (.024) .01 (.024) .006 (.024)
Outbreak × Married -.121⁎⁎⁎⁎ (.026) -.115⁎⁎⁎⁎ (.026) -.079⁎⁎ (.031)
Outbreak × Income -.001 (.001) -.001 (.001) -.001 (.001)
Constant .437⁎⁎⁎⁎ (.006) .436⁎⁎⁎⁎ (.006) .365⁎⁎⁎⁎ (.015) .361⁎⁎⁎⁎ (.016) .372⁎⁎⁎⁎ (.016) .347⁎⁎⁎⁎ (.017) .434⁎⁎⁎⁎ (.006) .346⁎⁎⁎⁎ (.017) .346⁎⁎⁎⁎ (.016) .327⁎⁎⁎⁎ (.018) .434⁎⁎⁎⁎ (.008) .348⁎⁎⁎⁎ (.018) .378⁎⁎⁎⁎ (.009) .341⁎⁎⁎⁎ (.017) .403⁎⁎⁎⁎ (.009) .344⁎⁎⁎⁎ (.018) .328⁎⁎⁎⁎ (.018)
Notes:
⁎ p ≤ .05;
⁎⁎ p ≤ .01;
⁎⁎⁎ p ≤ .005;
⁎⁎⁎⁎ p ≤ .001
Standard errors are shown in parentheses below coefficient estimates.
We also examined the main effects of the demographic and economic variables on emotional well-being. Marriage and income were the only two variables that had a consistent, significant effect on emotional well-being. Specifically, married people enjoyed a higher level of emotional well-being than unmarried ones and a higher income was associated with a higher level of emotional well-being. These results are consistent with psychological well-being patterns in other countries examined in prior research (Lucas and Schimmack, 2009, Wood et al., 1989; ).
Next, we analyzed the data collected during the outbreak. We ran a series of regressions with emotional well-being as the dependent variable, perceived knowledge, actual knowledge, whether the individual resided in Hubei, age, sex, marital status, income, and the interaction terms between the demographic variables and perceived knowledge as predictors (see Table 2 ). Across all regression models, participants’ perceived knowledge about coronavirus infection was a consistent, significant predictor of their emotional well-being. However, their actual knowledge was not a consistent predictor. In other words, people's perceived level of knowledge about coronavirus infection served as a stronger protector of their emotional well-being during the outbreak than the actual amount of knowledge they possessed.
Table 2 Perceived knowledge helped protect emotional well-being during the coronavirus epidemic
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Model 11 Model 12 Model 13 Model 14 Model 15 Model 16 Model 17
Perceived Knowledge .067⁎⁎⁎⁎ (.019) .068⁎⁎⁎⁎ (.019) .067⁎⁎⁎⁎ (.019) .067⁎⁎⁎⁎ (.019) .07⁎⁎⁎⁎ (.019) .069⁎⁎⁎⁎ (.02) .072⁎⁎⁎⁎ (.019) .073⁎⁎⁎⁎ (.02) .188⁎⁎ (.073) .193⁎⁎ (.073) .08⁎⁎⁎ (.026) .084⁎⁎⁎ (.027) .099⁎⁎⁎ (.033) .098⁎⁎⁎ (.033) .094⁎⁎⁎ (.033) .096⁎⁎⁎ (.033) .227⁎⁎⁎ (.08)
Actual Knowledge .019 (.011) .020 (.011) .022* (.011) .021* (.011) .022* (.011) .022* (.011) .020 (.011) .022* (.011) .021* (.011) .022* (.011) .019 (.011) .022* (.011) .02 (.011) .021* (.011) .019 (.011) .021* (.011) .021* (.011)
Hubei -.137* (.059) -.138* (.059) -.137* (.059) -.139* (.059) -.138* (.059) .295 (.4) .275 (.401) -.137* (.059) -.14* (.059) -.137* (.059) -.137* (.059) .271 (.402)
Age -.003* (.001) -.003* (.001) -.002 (.001) -.002 (.001) -.002 (.001) .011 (.008) .012 (.008) -.002 (.001) -.002 (.001) -.002 (.001) .012 (.01)
Sex .006 (.023) .007 (.023) .006 (.023) .005 (.023) .008 (.023) .123 (.154) .125 (.154) .007 (.023) .006 (.023) .138 (.155)
Married -.03 (.029) -.031 (.029) -.031 (.029) -.034 (.029) -.031 (.029) .123 (.161) .141 (.162) -.031 (.029) -.029 (.192)
Income .0003 (.001) .0003 (.001) .0003 (.001) .0003 (.001) .0003 (.001) .007 (.007) .007 (.007) .006 (.007)
PK × Hubei -.106 (.098) -.102 (.098) -.101 (.098)
PK × Age -.004 (.002) -.004 (.002) -.003 (.002)
PK × Sex -.028 (.038) -.03 (.038) -.033 (.038)
PK × Married -.043 (.04) -.044 (.04) -.001 (.048)
PK × Income -.002 (.002) -.002 (.002) -.001 (.002)
Constant -.211⁎⁎ (.083) -.212⁎⁎ (.083) -.127 (.092) -.131 (.093) -.149 (.094) -.148 (.094) -.228⁎⁎ (.084) -.164 (.096) -.608* (.293) -.639* (.295) -.269* (.111) -.206 (.12) -.31* (.132) -.258 (.139) -.319* (.134) -.253 (.14) -.776* (.322)
Notes:
⁎ p ≤ .05;
⁎⁎ p ≤ .01;
⁎⁎⁎ p ≤ .005;
⁎⁎⁎⁎ p ≤ .001
Standard errors are shown in parentheses below coefficient estimates.
We tested whether sense of control mediated the effect of perceived knowledge on emotional well-being. We ran a mediation analysis using a bootstrapping technique with 10,000 resamples (Model 4, Hayes, 2013). This analysis indicated that perceived knowledge had a significant positive effect on sense of control (a = .37, SE = .02, t = 20.49, p < .001) and that sense of control had a significant positive effect on emotional well-being (b = .23, SE = .02, t = 12.49, p < .001). Moreover, the otherwise significant direct effect of perceived knowledge on emotional well-being (c = .07, SE = .02, t = 3.55, p < .001) became non-significant (c’ = -.02, SE = .02, t = -.97, p = .33) after the indirect effect through sense of control was taken into account. The 95% bias corrected confidence interval for the indirect effect did not include 0 (95% CI = [.07, .10]), indicating a significant mediation. That is, sense of control mediated the relationship between perceived knowledge and emotional well-being.
As robustness checks, we reran the mediation analysis with emotional well-being as the dependent variable, perceived knowledge as the independent variable, actual knowledge as a covariate, and sense of control as the mediator. This analysis also yielded a significant indirect effect of perceived knowledge on emotional well-being through sense of control (95% CI [.07, .10]). We also reran the mediation analysis with actual knowledge, the demographic and economic variables and their interaction terms with perceived knowledge as covariates. This again yielded a significant indirect effect of perceived knowledge on emotional well-being through sense of control (95% CI [.02, .11]).
These mediation results provide evidence for our proposed psychological mechanism. That is, participants’ perceived knowledge about coronavirus infection was associated with a higher sense of control, which in turn protected their emotional well-being during the outbreak.