Using Divide-and-Conquer to Improve Tax Collection: Evidence from the Field

In the context of collecting property taxes from 13432 households in a district of Lima (Peru), we investigate whether prioritized enforcement can improve the effective use of limited enforcement capacity. We randomly assign households to two treatment arms: one replicating the city’s usual collection policy, and one implementing a prioritized enforcement rule in which households are ordered according to a suitable rank and sequentially issued clear short-term promises of collection if they fail to make minimum tax payments. Raw findings show that prioritized enforcement improved tax collection by increasing tax revenue, and decreasing the number of costly collection actions taken. We identify an important friction ignored by existing theory: tax-payers’ response to incentives is slow, which changes the optimal management of collection promises. Finally, we estimate a model of tax-payer behavior and use it to produce counterfactual treatment estimates for other collection policies of interest. In particular, we estimate that, keeping the number of collection actions fixed, prioritized enforcement would increase tax revenue over 5 months by 11.3%.