Experimental Evidence of Self-Serving Omission Bias

ANDREW HAYASHI, Davis Polk & Wardwell

Abstract

People evaluate outcomes, in part, by how those outcomes came about and who caused them. Causal attribution, however, is a process both complicated and subject to bias. I generate individual-level data from a variation on the Dictator Game in which the participants’ endowments are manipulated to identify one aspect of how people care about causal histories. The data are inconsistent with both preferences defined solely over outcomes and a general bias toward the status-quo/inaction. Subjects appear to care independently, but conditionally, about the effect of their own actions, and only demonstrate a bias toward inaction when it’s in their self-interest.