Perceived relative personal risk is low and perceived public risk is high
People’s estimate of relative personal risk was positively skewed, while their perception of public risk was negatively skewed. In particular, 35.6% of participants indicated they believed they were less likely to get COVID-19 relative to others of their same age and gender, while only 19.1% indicated they believed they were more likely to get COVID-19 relative to others their age and gender. Almost half (45.3%) indicated they believed they were as likely to get COVID-19 as others their age and gender (Fig. 2a). As for public risk, the vast majority (89.9%) indicated they believed COVID-19 posed high danger to the health of the human population (Fig. 2b).
We next examined which factors were related to participants’ risk estimations. To that end we ran two logistic regressions – one predicting perceived relative personal risk and one predicting perceived public risk. The independent factors included were demographics (age, gender, income, education, ethnicity, number of children), political orientation, restriction in the participant’s state, satisfaction with health insurance as well as sense of control measured using the Sense of Control Scale (Lachman & Weaver, 1998). This questionnaire examines people’s sense of mastery and perceived constraints. Interestingly, only two factors predicted both perception of personal and public risk: political orientation (perceived relative private risk: β = -0.465, p < 0.001, perceived public risk β = -1.366, p < 0.001, also see Bruine de Bruin et al., 2020 for the importance of political orientation on COVID-19 risk estimates) and a sense of control (perceived relative private risk: β = -0.28, p = 0.033, Fig. 3a, Table S1; perceived public risk: β = -0.546, p = 0.043, Fig. 3b, Table S2). In particular, both Republicans and those who scored high on the sense of control questionnaire tended to perceive personal, and public risk as low. In addition, males (β = 0.328, p = 0.007) and younger individuals (β = 0.014, p = 0.001) were all more likely to perceive their relative personal risk as low. No other factor was significantly associated with perceived public risk. Note, that all factors are included in one regression, thus the influence of one factor is controlled for when examining the other.
Perceived relative personal risk, but not perceived public risk, is related to participants’ happiness
We were interested whether and how perceived risk was related to peoples’ sense of happiness in times of crisis. We thus asked our participants to indicate “how happy are you with your life these days?” on a scale from least happy time of my life to most happy time of my life. About half (52%) of the participants rated their happiness as lower than other times in their life. Importantly, people who perceived their relative personal risk as lower than others their age and gender were more likely to indicate they were happier (Beta from a logistic regression model predicting relative happiness from perceived relative personal risk, perceived public risk and all demographic variables as controls β = -0.201, p = 0.03, Fig. 4a, b, Table S3). Perceived public risk on the other hand was not associated with happiness (Beta from the same model β = 0.036, p = 0.877, Fig. 4a, c). While the result is correlational, it is plausible that the belief that one is relatively immune to COVID-19 protected people’s sense of happiness during the crisis and/or that happiness alters personal risk perception.
Adding “sense of control” into the previous regression revealed that sense of control was the variable most strongly associated with happiness. The Beta coefficient signifying the magnitude of the relationship between sense of control and happiness (β = 0.99, p < 0.001) was more than double that of any other variable. In addition to a high sense of control those who were satisfied with health insurance (β = 0.43, p = 0.006), older individuals (β = 0.011, p = 0.025), those with higher income (β = 0.084, p = 0.011) and Republicans (β = 0.347, p = 0.023), all reported greater happiness (Fig. 4d, Table S4). We note that all these factors were included in one model, thus the results indicate independent effects of each variable.
It has been shown that sense of control is related to both optimism and happiness (Lachman & Weaver, 1998; Larson, 1989; Seligman, 2011). We thus examined if the relationship between perceived relative personal risk and happiness was mediated by participants’ sense of control. To that end, we computed a mediation model, controlling for all demographic variables (see Sect. 1). The model first confirmed that perceived relative personal risk was negatively related to happiness (total effect: β = -0.201, p = 0.03). Importantly, this relationship was mediated by the sense of control (indirect effect: β = -0.201, p = 0.045, Sobel Test: z = -2.009). Indeed, once the sense of control was statistically accounted for, the relationship between perceived relative personal risk and happiness was reduced to trend level (c’: β = -0.167, p = 0.08). In contrast, sense of control predicted happiness even when perceived relative personal risk was accounted for (path b: β = 0.99, p < 0.001). The reverse mediation was not significant. That is perceived relative personal risk did not mediate the relationship between sense of control and happiness (total effect: β = -0.90, p < 0.001; indirect effect: β = -0.034 p = 0.24, Sobel Test: z = 1.176; path c’: β = 0.99, p < 0.001; path a: β = -0.254, p = 0.051; path b: β = -0.167, p = 0.08). These findings suggest that a sense of control mediates its relationship with happiness, but not vice versa.
Perceived relative personal risk and perceived public risk are related to anxiety
Thus far we have shown that perceived relative personal risk, but not perceived public risk, was associated with happiness during the COVID-19 crisis. We next examined if these factors were related to anxiety, which we measured using the short version of STAI (Marteau & Bekker, 1992). We found that perceived risk, both personal and public, were strongly associated with high anxiety (Betas from a logistic regression including all demographics, perceived relative personal risk: β = 0.325, p = 0.001, Fig. 5a, b, perceived public risk: β = 0.681, p = 0.004, Fig. 5a, c, Table S5). This was true also when adding sense of control into the model, which in itself was negatively associated with anxiety and was the strongest predictor of it (β = -0.854, p < 0.001). In addition, younger individuals, females, Caucasians, and Democrats were more anxious (age: β = -0.027, p < 0.001; gender: β = 0.505, p < 0.001, ethnicity: β = 0.582, p = 0.004, political orientation: β = -0.379, p = 0.016, Fig. 5d, Table S6).
To examine if the relationship between perceived relative personal risk and anxiety was mediated by participants’ sense of control, we computed a mediation model, controlling for all demographic variables (see Sect. 1). Once again, perceived relative personal risk was positively related to anxiety (total effect: β = 0.325, p < 0.001). This relationship was partially, but not fully, mediated by a sense of control (indirect effect: β = 0.173, p = 0.05, Sobel Test: z = 1.9470), as after accounting for a sense of control perceived relative personal risk was still related to anxiety (c’: β = 0.303, p = 0.003). By contrast, we did not find evidence for the reverse mediation. That is perceived relative personal risk did not mediate the relationship between sense of control and anxiety (total effect: β = -0.876, p < 0.001; path c’: β = -0.854, p < 0.001; indirect effect: 0.062, p = 0.163, Sobel Test: z = 1.395, path a: β = -0.254, p = 0.051; path b: β = 0.302, p = 0.003). Sense of control did not mediate the relationship between perceived public risk and anxiety, although a trend was observed (indirect effect: 0.426, p = 0.083, Sobel Test: z = 1.732) which became significant in Time 2 (see Online Appendix).
Perceived public risk, but not perceived relative personal risk, is associated with behavioral compliance
The above results show that perceived risk is associate with people’s emotional state during the pandemic. We next examined whether it is also associated with people’s behavioral response to it. To that end, we assessed participants’ self-reported behavioral compliance with government officials’ advice to mitigate the COVID-19 outbreak. We found that behavioral compliance was high: 97% of participants reported putting effort into social distancing, 95% reported putting effort into frequent hand washing and 77% into avoiding face touching, 95% reported they did not visit other people’s homes in the last week and 82% reported they have not come within 1 m of people outside their own residence.
We averaged participants’ raw scores on all these measures such that each participant had one score reflecting behavioral compliance. Following the guidelines of the Journal of Risk and uncertainty this variable was binarized with 0 summarizing responses below the midpoint of the scale (< 50) and 1 summarizing responses equal or above midpoint of the scale. These scores were then entered as a dependent variable in a binomial logistic regression model. We found that while perceived public risk was strongly related with behavioral compliance (Beta from a logistic regression model including perceived relative personal risk, perceived public risk and all demographic controls: β = 2.03, p < 0.001, Fig. 6a, c, Table S7), perceived relative personal risk was not (β = -0.272, p = 0.362, Fig. 6a, b). Adding sense of control into the model did not alter the results (β = 1.946, p < 0.001), which in itself was not related to behavioral compliance (β = -0.725, p = 0.203, Fig. 6d, Table S8). The relationship between perceived public risk and behavioral compliance could not be explained by high anxiety alone, as even when we add anxiety into the model, the effect of perceived public risk on behavioral compliance remains significant (Betas in a model including perceived relative personal risk, perceived public risk, anxiety and all demographic controls Anxiety: β = 0.429, p = 0.34, Perceived public risk: β = 1.98, p < 0.001, Table S9).