Public auditors should reduce agency problems and improve transparency. We address the question of whether auditors should be elected by the citizens or appointed by either the legislature or the executive, and explore the influence of conducting performance audits. We construct a unique dataset at the US State level capturing differences in the institutional design of state auditing institutions. We estimate the influence of auditor characteristics on different outcome variables reflecting government performance and implement an alternative identification strategy relying on citizens' electoral decisions. We examine whether citizens use divided government - a costly mechanism to control the government - as a substitute, when other effective, but less costly mechanisms are not available. Even if the empirical results are sometimes difficult to interpret, we generally find that (1) performance audits tend to be beneficial and (2) elected auditors with a strong mandate to conduct performance audits seem to outperform other institutional arrangements.
