Choice Architecture for Messaging Corruption

• In Event: Experiments in Information and Accountability

Abstract

The literature on corruption and responsibility attribution suggests that voters update when informed about malfeasance. Evidence is ambiguous. Avis (2018) find that elected official malfeasance responds to audit results but this is unlikely a function of the electoral mechanism. Ferraz (2011) find that malfeasance is lower for elected officials that face re-election constraints. An important mechanism in question here is the updating of beliefs by the average voter when they are informed about these audit metrics.

We conducted a 3x4 factorial design to test whether voters update their beliefs about the levels of malfeasance of their local government to different information frames about objective measures of corruption. Respondents received a 40-second video via WhatsApp with information about the levels of malfeasance. It compared their municipal government to either 1) the nearby municipalities (spatial frame); 2) the same municipality with respect to its previous audit (temporal frame); 3) and to all municipalities in the country (spatial frame 2). Respondents were randomly assigned to 1 of the 4 levels of the information content factor, where the content highlighted either 1) the severity of the irregularities; 2) the total resources misappropriated; 3) The percentage of resources involved in corruption with respect to the total municipal budget; 4) And the programs that were audited. We measured beliefs and their saliency about the prevalence of corruption.

Authors

• Raymond Duch, CESS Nuffield

• Felipe Torres, University of Oxford