Figure 1.

Outcome Means by Treatment Arm

Notes: Figure shows the mean of each primary outcome by treatment condition among the sample of Black respondents (dark-blue bars) as well as among the sample of White respondents (light-blue bars). Sender and signal rating are inverse-covariance-weighted indices as described in Anderson (2008); flu and COVID-19 vaccine intent have support 0 to 1. For dark-blue bars, p-values test the null hypotheses that the concordant expert, concordant nonexpert (standard signal condition), and discordant expert (acknowledgement condition) means each differ from the discordant expert (standard signal condition) among Black respondents. For light-blue bars, p-values test the null hypothesis that the concordant expert (standard signal condition) mean differs from the discordant expert (standard signal condition) among White respondents. Ninety-five percent confidence intervals using robust standard errors are shown.