This paper explores the role of identity in voters’ decision to retain corrupt
politicians. We build up a model of electoral accountability with pure moral hazard
and bring it to the lab. Politicians must decide whether to invest in a public
project with uncertain returns or to keep the funds for themselves. Voters observe
the outcome of the project but not the action of the politician; if the project is
unsuccessful, they do not know whether it was because of bad luck or because
the politician embezzled the funds. We run two treatments; a control treatment
and a treatment where subjects are assigned an identity using the minimal group
paradigm. Our main result is that, upon observing a failed project, voters approve
politicians of their same identity group significantly more often than in the control
and compared to politicians of a different group. This is partially driven by a belief
on same-identity politicians being more honest. We also observe that subjects acting
as politicians are much more honest than expected by the equilibrium prediction.