Avaliação da Curva de Juros Empregando Extensões do Modelo de Diebold & Li com Três Fatores

This paper evaluates methods that employ Kalman Filter to estimateDiebold and Li (2006) extensions in a state-space representation, applyingthe Nelson and Siegel (1987) function as measure equation and differentspecifications for the transition equation that determines level, slope andcurvature dynamics. The models that were analyzed have the followingstructures in transition equation: (1) AR(1) specification, employing adiagonal covariance matrix for the residuals; (2) VAR(1) specification,employing a covariance matrix calculated with Cholesky decomposition; (3)VAR(1) extension, inserting variables related to the Covered Interest RateParity (CIRP); (4) VAR(1) extension, including stochastic volatilitycomponents. The major findings of this paper were: (1) evaluating the latentvariables dynamics, the curvature was the factor that fitted better to thestochastic volatility component; (2) in a broad sense, even though thesimplest VAR(1) model was the one that provided the best out-of-sampleperformance in the most part of maturities and forecasting horizons, theextension inserting variables related to the CIRP was able to overcome theformer specification in some of these simulations.