Três ensaios sobre o gasto local no Brasil: descentralização, eficiência e voto
Título da Revista
ISSN da Revista
Título de Volume
In the first essay this dissertation starts evaluating the world context of fiscal decentralization and democratization that Brazil was involved during the last years of the 20th century. In an empirical analysis for the developing world, the first research stresses the influence of the type of regime in the relationship between fiscal decentralization and government size. The estimations, using system‐GMM, show that there is a threshold of fiscal decentralization in the range of 20% and 30%.Democracies from the developing world that decentralized their fiscal resources above this threshold obtained government sizes smaller than developing countries with dictatorial regimes. These results, which emphasize the role of democracy and local government in public expenditures, led the sequence of the work to analyze the public expenditure efficiency of Brazilian municipalities and its implication in the voting process. In the second essay, measures of productivity and efficiency change are calculated (Malmquist Index) for the municipal investments in health and education, between the period of 2004 and 2008. The results from stochastic frontier analysis show an improvement in the production frontier for both areas (an 18.7% average progress in TFPC for investments in education and a 14.2% for health outlays). Most of the advancements were derived from developments in Technical Change, instead of growth in Technical Efficiency Change. In the third essay the measures of efficiency and productivity change are used to question if voters in municipal elections reward incumbents that improved the efficiency of public investments in health and/or education. The last essay’s results indicate that the answer is positive for improvements in education expenditures but negative for advances in health expenses. In order to correct likely measurement errors of the productivity coefficients instrumental variables are used in 2SLS regressions.