Table 2.
Main effects for COVID-19 behavior engagement as a function of demographics, affect, impulsivity, news exposure, norms, and perceived behavior effectiveness.
| Indicator. | Social Distancing | Indoor Mask Use | Outdoor Mask Use | |||
|---|---|---|---|---|---|---|
| β | R2 | β | R2 | β | R2 | |
| Covariate Indicators | ||||||
| Age | −0.01 | 0.00 | −0.05 | 0.00 | 0.01 | 0.00 |
| Race/ethnicity | −0.10 | 0.01 | −0.02 | 0.00 | −0.16 ** | 0.03 |
| Sex | 0.08 | 0.01 | 0.33 *** | 0.11 | 0.07 | 0.00 |
| State Response Timing | −0.15 ** | 0.02 | −0.26 *** | 0.07 | −0.31 *** | 0.10 |
| Covariate Model | 0.04 | 0.16 | 0.13 | |||
| Affect and Impulsivity Indicators | ||||||
| GAD-7 | 0.11 * | 0.01 | 0.14 | 0.02 | 0.11* | 0.01 |
| PANAS-C Negative | 0.12 ** | 0.01 | 0.06 | 0.00 | 0.13 ** | 0.02 |
| PANAS-C Positive | −0.17 *** | 0.03 | −0.05 | 0.00 | −0.08 | 0.01 |
| Negative Urgency | −0.06 | 0.00 | −0.04 | 0.00 | −0.09 | 0.01 |
| Positive Urgency | −0.04 | 0.00 | −0.16 * | 0.03 | −0.07 | 0.00 |
| Sensation Seeking | −0.10 | 0.01 | −0.09 | 0.01 | −0.03 | 0.00 |
| COVID-19 Indicators | ||||||
| COVID News Watch | 0.19 *** | 0.04 | 0.05 | 0.00 | 0.23 *** | 0.05 |
| COVID News Search | 0.16 *** | 0.03 | 0.06 | 0.00 | 0.25 *** | 0.06 |
| Indoor Mask Norms | −0.01 | 0.00 | 0.41 *** | 0.17 | 0.07 | 0.00 |
| Outdoor Mask Norms | 0.12 ** | 0.01 | 0.23 ** | 0.05 | 0.49 *** | 0.24 |
| Indoor Mask Perceived Effectiveness | 0.29 *** | 0.08 | 0.45 *** | 0.20 | 0.38 *** | 0.14 |
| Outdoor Mask Perceived Effectiveness | 0.31 *** | 0.10 | 0.38 *** | 0.14 | 0.61 *** | 0.37 |
Variables were coded such that non-White = 0 and White = 1; male = 0 and female = 1; early state response timing = 0 and late state response timing = 1. Early response is defined as a stay-at-home order before 1 April 2020. Late response is defined as a stay-at-home order on 1 April 2020 or later, including states that never implemented an order. Each main effect model was conducted individually (N = 476–490). Models reported in this table controlled for age, race, sex, and response timing. Given the number of models tested, a false discovery rate correction was applied to determine significance. Significant models after applying the correction are denoted as follows: * = p < 0.05, ** = p < 0.01, *** = p < 0.001. Analyses were also conducted using the PROC MIXED procedures in SAS for multilevel modeling with university as the nesting variable to account for random variance that may be attributable to university affiliates. Results were generally consistent with the linear regression models presented in Table 2, and no significant interactions emerged between university affiliation and the independent variables. Using multilevel modeling, state response timing was no longer a significant indicator of any outcome despite moderate to large Betas (βs = −0.59 to 0.80). Additionally, multilevel models examining relations between state response timing and subsequent outcomes resulted in large standard errors (values ranged from 0.48 to 0.49). This is likely due to the fact 206 of 209 participants in late response states were affiliated with the same university. As such, there is multicollinearity between the university affiliation and state response timing that is masking the significance of this variable and resulting in inflated standard errors. Given that models were largely unchanged and the present issues with multicollinearity between university and state response timing, the results derived from linear regressions are presented).