Modelo FAVAR Canônico para Previsão do Mercado de Crédito

This study proposes a new methodology called Canonical FAVAR that incorporates the canonical correlation analysis in the estimation of two-step FAVAR models to obtain more appropriate factors for forecasting. The canonical correlation technique identifies a small number of linear combinations of major components that have better correlation with the variables of interest and therefore greater predictive ability. The Canonical FAVAR was applied in forecasting the Brazilian financial system credit variables and their predictive ability was compared to one and two - step FAVAR models. These models were adjusted for five variables of the credit market, and the results were better than those obtained by traditional FAVAR