Brazil opted for decentralization in the provision of health services at the time of the 1988 Constitution´s promulgation. The municipalities became entirely responsible for primary healthcare provision, sharing financing responsibilities with the central government. The aim of this paper is to evaluate the impact of health spending autonomy on infant mortality rates, using Brazilian local data from 2000 to 2007. As a measure of spending autonomy we used the share (%) of overall health expenditures financed by municipalities’ own resources. The larger the share of health expenditures which is not financed by the central government (grants), the greater the autonomy, because local governments can decide how and on what to spend these resources, and most of intergovernmental transfers to the health sector are allocated to predefined specific purposes. The fixed effects estimates, through the use of an instrumental variables approach, show that, on average, greater spending autonomy is not associated with lower infant mortality rates. However, given that local governments and populations are different in many aspects, it is reasonable to expect a heterogeneous response to decentralization. Galiani et al. (2008), for instance, assess the hypothesis that decentralization of high schools in Argentina may have increased the difference in educational outcomes. Could the same have happened to the quality of health in Brazilian municipalities? In well-managed municipalities and in which individuals participate actively and can support their preferences, fiscal autonomy ought to improve the delivery of public services. Using a measure of efficiency in the delivery of health services to split the municipalities in two groups, we got significant results for the more efficient municipalities, implying an increase in inequality. This measure corresponds to efficiency scores estimated using data envelopment analysis (DEA). We also test for heterogeneous impacts of the Family Health Program (PSF), the most important Brazilian basic health care policy, which main expansion occurred until 2007. It is a federal Program that each municipality runs independently. Again, we find evidence of a heterogeneous response, since PSF only reduces infant mortality rates in efficient municipalities. We then confirm Galiani’s (2008) hypothesis that decentralization may help only the best to do even better.