Are socio-demographic and economic characteristics good predictors of misinformation during an epidemic?

We combine data on beliefs about the origin of the 2014 Ebola outbreak with two supervised machine learning methods to predict who is more likely to be misinformed. Contrary to popular beliefs, we uncover that, socio-demographic and economic indicators play a minor role in predicting those who are misinformed: misinformed individuals are not any poorer, older, less educated, more economically distressed, more rural, or ethnically different than individuals who are informed. However, they are more likely to report high levels of distrust, especially towards governmental institutions. By distinguishing between types of beliefs, distrust in the central government is the primary predictor of individuals assigning a political origin to the epidemic, while Muslim religion is the most important predictor of whether the individual assigns a supernatural origin. Instead, educational level has a markedly higher importance for ethnic beliefs. Taken together, the results highlight that government trust might play the most important role in reducing misinformation during epidemics.