Modelos para previsão do índice de volume trimestral do PIB do estado de Minas Gerais

Abstract

This monograph mainly aims at the construction of statistical models for monitoring and forecasting the quarterly series of volume index of gross domestic product (GDP) of Minas Gerais, traditionally published by Fundação João Pinheiro. The strategy adopted was to use its own series (univariate) or other concomitant variables that explain the variable of interest (multivariate models). In the first case, the purpose has been achieved with the use of typical approach for time-series Box-Jenkins (ARIMA models or the more general case - SARIMA models) and through exponential smoothing technique Holt-Winters. In the second case, an analysis was done via the construction of a dynamic regression model. The monograph also justifies the importance of modeling (and try to predict) such a complex and variable and attempts to fill a gap in relation to the behavior of the State economy. Finally, it presents the main conclusions obtained identifying which model performed best in the sample period and the validation period.