A COMPARISON OF THE PREDICTIVE POWER OF DIFFERENT VARIABLES FOR PERIODS OF RECESSION IN THE BRAZILIAN ECONOMY

Abstract

This paper seeks to compare the predictive power of different variables during periods of recession for the Brazilian economy from 1997 until 2017. The econometric model used is the simple probit that provides the probability of the economy
going into recession from one to eight quarters ahead. The model makes it possible to
estimate the pseudo R2 as a parameter of comparison between the predictive power of
the variables analyzed. Among the results found, the use of variables that capture the
country’s fiscal situation, the international commodity price index and financial variables, such as the return of the Ibovespa index and the country risk stand out for shortterm forecasts. However, this article identifies a gap in the Brazilian literature regarding
variables that can forecast recessions for more distant time horizons.