Um procedimento para prever recessões no Brasil a partir de indicadores antecedentes ^♦

Abstract

The goal of this paper is to test for Brazil a new recession forecasting approach that has been proposed by authors from other research fields, such as physics, which can be useful for predicting extreme events also in economy. From the unemployment series (1985-2012) smoothed by splines, we first identify through fractal analysis variables that present co-movement along the period of analysis. Next, we used discriminant analysis to identify leading indicators for the unemployment series. We concluded that unemployment periods are in general preceded by an one year lagged improvement in terms of trade, rises in imports, and by decreases in minimum real wages measured in terms of in PPP.

Keywords
Extreme events; Forecast; Fractal analysis; Discriminant analysis