Spillovers in a decentralized health economy

Abstract

We estimate the direct as well as the spillover effect of a federal grant (Municipalities’ Participation Fund, FPM)) on local health indicators in Brazilian municipalities. We use a Regression Discontinuity Design (RDD) when exploring the FPM distribution rule according to population brackets, and we disentangle each effect considering neighbouring towns near different thresholds. Our quasi-experimental estimates show that the FPM spillovers improve local health indicators but reduce the provision of public goods, especially when the neighbouring municipality receiving the additional transfer is small. In particular, our results show that the decentralization of health services could lead to an under-provision of health services in terms of the number of doctors (-0.35% and -0.87%, respectively, for cities with smaller populations), especially general practitioners and surgeons (-1.84% and -2.45%, for the most populous cities in our sample). At the same time, the direct effect is positive as expected, particularly in the Family Health Programme — the main preventive programme in small towns — where there is an increase in PSF visits (1.59%) and PSF visits with a doctor (1.8%) or a nurse (2%). We also find positive effects on hospitalization and complex services in the major cities in the sample and reductions in the infant mortality rate (-0.18%) and morbidity rate (-0.41%). The direct impacts are reduced when we control for neighbours’ FPM, which shows that the spillover effects and spatial interactions are important for explaining the FPM effect on health outcomes. We test whether the negative spillovers are caused by the lack of policy coordination among neighbouring cities, aside from the reduction in regional demand for health services. We find that the spillover effects are stronger when there is more competition in mayoral elections and when neighbouring mayors are not from the same party, which shows that political incentives are important for explaining the observed spillovers.