Belief elicitation in political protest experiments: When the mode does not teach us about incentives to protest

Many recent experiments studying political protests elicit subjects' beliefs and actions, before and after an information intervention, to assess the causal role of beliefs on actions. We show that unless beliefs are symmetric and unimodal, using a belief elicitation scheme that is mismatched with the research question may affect the magnitude and even reverse the sign of identified effects. We provide a simple characterization of when such a sign reversal occurs. As an example, we revisit Cantoni et al. (2019)'s influential study of whether political protests are strategic complements or substitutes. We show how their belief elicitation method allows, in theory, a novel and alternative interpretation of their results, which could have been avoided with a different method.