The structure of the Australian growth process: A Bayesian model selection view of Markov switching

In this paper we examine the structure of the GDP growth process in Australia. Our objective is to determine whether the growth process exhibits well-defined business cycle features and whether any significant regime shifts have occurred in the mean and variance of the growth process, as well as its autoregressive structure. The analysis is based on a general Markov Switching Autoregressive Model, implemented from a Bayesian perspective using Markov Chain Monte Carlo numerical methods. Our results suggest that there are two regimes in GDP growth and that the dominant statistical feature is a switch from higher to lower volatility in growth. In contrast to previous studies for Australia, our results suggest that there are no distinct business cycle features present in the growth process.