Published online by Cambridge University Press: 12 January 2015
I investigate the extent to which reputational incentives affect policy choices in the context of a controlled laboratory experiment. In theory, asymmetric information and outcome unobservability undermine electoral delegation by creating incentives for politicians to pander. Under the right conditions, it may be preferable to remove such incentives by removing accountability altogether. The data suggest that subjects playing the role of politicians fail to take advantage of voters even though voters indeed create the predicted electoral incentives, albeit in a weaker form than predicted by the theory. When given the choice of institutions via a novel elicitation method, subjects prefer to retain electoral accountability or to make decisions themselves through direct democracy, even though both institutions yield lower expected payoffs than delegation to unaccountable agents.