Table A2.

Robustness to Omitted Variable Bias.

	(1)	(2)	(3)	(4)	(5)	(6)
Dependent variable:	Trust in scientists	Scientists working for private companies benefit the public	Scientists working for private companies are honest	Scientists working for universities benefit the public	Scientists working for universities are honest	Scientists to find out accurate information
Exposure to Epidemic (18–25)	−3.454**	−1.283***	−1.731***	−0.616	−3.330***	−1.438**
	(1.330)	(0.338)	(0.642)	(0.478)	(0.446)	(0.664)
Bounds on the treatment effect (δ = 1, Rmax = 1.3*R)	(-3.454, −3.044)	(-1.238, −1.134)	(-1.731, −2.301)	(-0.616, 0.649)	(-3.330, −2.587)	(-1.438, −0.827)
Treatment effect excludes 0	Yes	Yes	Yes	Yes	Yes	Yes
Delta (Rmax = 1.3*R)	−44.41	−132.15	13.52	−1.97	−39.39	−5.367

Notes: *** Significant at the 1% level; ** Significant at the 5% level; * Significant at the 10% level. Bounds on the Democracy 18–25 effect are calculated using Stata code psacalc, which calculates estimates of treatment effects and relative degree of selection in linear models as proposed in Oster (2019). Delta, δ, calculates an estimate of the proportional degree of selection given a maximum value of the R-squared. Delta is assumed to be 1 in the analysis, which means that the observed and the unobserved factors have an equally important effect on the coefficient of interest. Rmax specifies the maximum R-squared which would result if all unobservables were included in the regression. We define Rmax upper bound as 1.3 times the R-squared from the main specification that controls for all observables.