Previsão da Curva de Juros: um modelo estatístico com variáveis macroeconômicas

A variety of models has been proposed for yield curve forecasting. In this paper we present a dynamic latent factor model for Brazilian interest rate term-structure forecasting, based in three major information sources: macroeconomic variables, surveys and risk premium. We use the proposed model to produce forecasts six month ahead and we compare the results with the well known Diebold and Li (2006) and a random walk. Our forecasts appear much more accurate than the alternative models.