Appendix 1—table 1. Robustness of main paper results to alternative regression types and the inclusion/exclusion of covariates.

	Logistic regression models with results presented as average marginal effects		Linear probability models
	With covariates (main paper results)	Without covariates	With covariates	Without covariates
Effect of labeled risk compared to unlabeled risk	3.0 pp (p=0.003)	3.0 pp (p=0.003)	3.0 pp (p=0.003)	3.0 pp (p=0.003)
Effect of comparison to motor-vehicle mortality compared to no comparison	2.4 pp (p=0.049)	2.4 pp (p=0.051)	2.4 pp (p=0.0498)	2.4 pp (p=0.052)
Effect of comparison to COVID-19 mortality compared to no comparison	0.8 pp (p=0.496)	0.7 pp (p=0.568)	0.8 pp (p=0.497)	0.7 pp (p=0.571)
Effect of relative comparison compared to absolute comparison	1.3 pp (p=0.285)	1.3 pp (p=0.277)	1.3 pp (p=0.286)	1.3 pp (p=0.277)
Effect of both risk labeling and a motor-vehicle comparison compared to no labeling or comparison	6.1 pp (p<0.001)	6.1 pp (p<0.001)	6.1 pp (p<0.001)	6.1 pp (p<0.001)

Notes: Outcome: ‘Would you take this vaccine?’ (yes = 1, others = 0). As per our pre-analysis plan, covariates include age, sex, education, and country. Sample sizes for the relative to absolute comparison (N=5998) and effect of both labeling and a motor-vehicle mortality comparison (N=3002) are smaller since they are only estimated among subset of the total sample.