Bayesian Inference Applied to Dynamic Nelson-Siegel Model with Stochastic Volatility

Abstract

In this article we propose to estimate a dynamic model for the term structure of interest rates considering two alternative specifications of the Nelson and Siegel (1987) model, generalizing several models found in literature. At first, we consider the weights of factors time-varying with conditional heteroskedasticity using a stochastic volatility model with common factors. In the second case, we consider a model where the latent factors individually follow autoregressive processes with stochastic volatility, including the possibility of leverage effects. The so-called volatility factorsseek to capture the uncertainty over time associated with the level, slope and curvature of the yield curve. The estimation is performed using Bayesian inference using Monte Carlo Markov Chain methods. The results for the term structure of DI futures and U.S. Treasuries used in this studyshow that the volatility factors are highly persistent, and also indicate that the use of stochastic volatility lead to better in-sample fits to the observed yield curve.