Leo Czajka

IRES/LIDAM, UCLouvain

will give a presentation on

Fraud Detection under Limited State Capacity: Experimental Evidence from Senegal

Abstract: Information on taxpayers is essential to efficiently allocate tax administrations' scarce resources in low income countries, but it is complex to collect. In the context of Senegal, this paper shows that a randomized low-cost communication campaign can induce some of the largest firms to submit more complete third-party reports about their service suppliers. The intervention increases the prevalence of suppliers' numerical ID in third-party reports by 52% on average. We estimate this improves identification information for about 18,546 client-suppliers pairs in total, thereby facilitating the tracking of $1.6 billion, i.e the equivalent of 6 % of GDP. Matching suppliers' ID to the tax registry and the universe of tax declaration we show that the new information increases the probability to detect payments going to under-reporting formal suppliers by 44%. A simple simulation exercise shows that exploiting the newly available information to strategically audit the largest under-reporting formal suppliers could increase audit returns by 115%.