Assessing macro influence on Brazilian yield curve with affine models

Using affine term structure models with latent and observable factors, we study the interaction between macro variables and the Brazilian yield curve contained in a daily dataset. Two versions of the model are tested, one in continuous time and estimated using maximum likelihood with Chen-Scott inversion (1993), and the other in discrete time and estimated using Monte Carlo-Markov Chain (MCMC) with Kalman filter. The effect of the inclusion of macro factors and changes of specification on the goodness-of-fit and dynamics are analysed. We conclude that the interaction between the macro variables and the yield curve is an important element for understanding the dynamics. Also, the choice of specification alters the response of the identified shocks.