Bayesian Estimation of Common Factor Models with Long-range Dependence

Abstract

We present a sampling-based Bayesian approach for modeling and forecasting a common factor time series model in which the common components are long-range dependent. The Gibbs sampling framework allows us to use a less computationally demanding ARMA model to approximate the common long-range dependent behavior in the sampling algorithm; we then adjust for the approximation using importance sampling. The sampling framework also allows for easy incorporation of other features, such as possible outlying observations and non-Gaussian disturbances. We illustrate the estimation procedure using simulated data and real inflation data for two neighboring European countries. We then discuss estimation of a non-Gaussian multivariate stochastic volatility (MVSV) model with common long- range dependent components in the Bayesian framework. The common component MVSV model is applied to stock return volatility data for companies having similar annual sales.