Regime Switching Models for Gas Prices

Abstract

This thesis is mainly concerned with a two-state regime switching model with two Ornstein-Uhlenbeck processes and its application to modelling natural gas prices. We start from analyzing Hamilton's (2005) model, a two-state regime switching model in a discrete time setting, and his recursive filtering approach to parameter calibration. Then we discuss the efficiency of models by adding seasonality to the long term mean. We also develop a recursive Bayesian filtering approach and compare it with Hamilton's (2005) filtering approach. This Bayesian calibration approach is applicable to regime switching models in both discrete and continuous time settings.