To improve service delivery public sector organizations often rely on reports from interested parties
that are costly to verify. If accurate, this information can then serve the dual purpose of incentivizing
performance and reducing policy mistakes. Received wisdom suggests that the benefit of raising
performance via financial incentives should be balanced against the risk of collusion between agent
and monitor, resulting in false reporting. We evaluate two different forms of local monitoring by
head-teachers of Ugandan primary schools, randomly varying whether reports of teacher attendance
trigger financial incentive payments or not. A theoretical model provides a normative framework to
make welfare comparisons, taking into account teacher attendance, the cost of monitoring, as well as
policy mistakes due to false reporting. Consistent with the model, we find that teacher attendance,
monitoring frequency, and the number of false reports all increased with the introduction of financial
incentives. More surprisingly, but again consistent with the theory, we find that the number of policy
mistakes actually decreases with incentives: there were more false reports but this effect was counterbalanced
by more reports in general and (hence) fewer mistakes caused by a lack of information. We
combine empirical estimates with the theoretical framework to bound welfare gains from attaching
financial incentives to local reports, and conclude that welfare was higher with financial incentives.