Comparative Analysis of the Allocative Efficiency of Municipal Public Expenditures using Data Mining

Research purpose: Analyze the allocative efficiency of public expenditures of Brazilian municipalities using data mining techniques. The municipalities considered efficient were those that had the best relation between their expenses and their results in the socioeconomic indicators, that is lower expenses and better results when compared to other cities. Theoretical framework: Theories of the Federation, specifically Fiscal Federalism, in which one of the main propositions is that decentralized public taxes and expenses are more efficient. Methodology: Collection and integration of financial and operational data of the years 2010, 2011, 2015, 2016 and 2017. Transformation into indicators. Cleaning and selection of indicators for transformation into categorical attributes (discretization). Results: Feasibility and importance of using data mining over comprehensive data sets. Generation of decision trees, based on data mining algorithms, allowing the identification of common characteristics of municipalities with outstanding performance. Relevant patterns were identified in the areas of education, health and development, such as the importance of spending resources on teacher training and characterized regional discrepancies in terms of the performance of the Brazilian Unified Health System (SUS). Originality: The study fills the gap in the use of Data Mining techniques in national works, aimed at comparing cities and respecting the reality in Brasil. In previous studies, it was possible to observe a repetition of the Data Envelopment Analysis (DEA) methodology for studying the efficiency of public expenditures. Theoretical and practical contributions: Proposition of a methodology for studying the efficiency of municipal public expenditure, valid from a theoretical and practical point of view, in multiple areas of activity, with the techniques and tools of Data Mining.