Misdiagnosis, Learning, and the Adoption of New Malaria Therapy in Tanzania

Abstract

We examine the role of social learning in the adoption of a new and effective therapy for malaria. In our model, individuals learn about the effectiveness of the new therapy by observing the health outcomes of past adopters. We show that misdiagnosis of malaria -- which is common in resource-poor settings -- can slow the learning process and stifle adoption. Using data from a pilot program which distributed the new therapy through health facilities in Tanzania, we estimate a learning effect and test for the role of misdiagnosis in learning and adoption. Our empirical strategy accounts for various potential sources of bias, including persistence in the local disease environment and selection into adoption based on the severity of illness. We find that when past adopters have poor health outcomes, the subsequent probability of adoption decreases. This learning effect is smaller in villages where misdiagnosis is more prevalent, and adoption rates in these places are lower over time. Finally, we simulate pairing the introduction of the new therapy with a new diagnostic test for malaria. The results of the simulation show that when misdiagnosis is reduced, learning occurs more quickly and adoption rates are higher.