Combinação de previsões: O caso do ICMS no Estado de Minas Gerais

Abstract

A major determinant for the implementation of public policies by public agencies is determining the availability of resources. In the case of the Brazilian States, are noteworthy the collections of the VAT. The objective of this monograph is to use the combination of forecasts as a way of improving the accuracy of forecasts of VAT for the State of Minas Gerais obtained by univariate time series models. For this, forecasts of SARIMA models and exponential smoothing were made which were later combined by five different procedures, namely; optimum combination, linear regression, past performance, simple mean and median. The results obtained indicate that the combination of forecasts, for whatevercombination procedure used, can improve the prediction performance significantly.