Table 2.

Results from final trimmed models (estimated separately for concurrent and antecedent predictors)

Perceived stress	Internalizing symptoms	Anger
Demographics and concurrent correlates •Hopelessness (β = 0.21, p < 0.001) •Lifestyle disruptions (β = 0.20, p < 0.001) •Economic disruption (β = 0.09, p = 0.009) •Actual illness – loved ones (β = 0.08, p = 0.014) •Migration background (β = 0.08, p = 0.035) •Health risks – loved ones (β = 0.06, p = 0.068)	Demographics and concurrent correlates •Economic disruption (β = 0.20, p < 0.001) •Lifestyle disruptions (β = 0.16, p < 0.001) •Hopelessness (β = 0.14, p < 0.001)	Demographics and concurrent correlates •Lifestyle disruptions (β = 0.21, p < 0.001) •Hopelessness (β = 0.14, p < 0.001) •Low trust in society's response (β = 0.12, p < 0.001) •Economic disruption (β = 0.11, p = 0.003) •Frequent COVID-19 news-seeking (β = 0.09, p = 0.014) •Female (β = 0.09, p = 0.011) •Health risks – self (β = 0.06, p = 0.082) •Medium education level (β = −0.06, p = 0.084, ref = high education)
Antecedent predictors •Previous perceived stress (β = 0.34, p < 0.001) •Stressful life events (β = 0.10, p = 0.003) •Bullying victimization (β = 0.09, p = 0.009) •Low self-rated health (β = 0.08, p = 0.018)	Antecedent predictors •Previous internalizing (β = 0.47, p < 0.001) •Bullying victimization (β = 0.10, p = 0.007) •Stressful life events (β = 0.08, p = 0.035)	Antecedent predictors •Previous anger (β = 0.22, p < 0.001) •Low generalized trust (β = 0.11, p = 0.004) •Perceived social exclusion (β = 0.10, p = 0.012) •Stressful life events (β = 0.09, p = 0.011) •Low self-rated health (β = 0.08, p = 0.036)

Note: Each final model included all demographic variables and the respective outcome at the previous assessment. For the upper row of results, all significant concurrent correlates from Fig. 2 were entered simultaneously; those with p ⩾ 0.10 were trimmed. For the lower row of results, all significant antecedent correlates from Fig. 2 were entered simultaneously; those with p ⩾ 0.10 were trimmed. Results for demographic factors and previous emotional distress are shown only once to avoid redundancy. Concurrent and antecedent predictors are ordered by size of the standardized regression coefficient.