Long-memory dynamics in a SETAR model - applications to stock markets

This paper presents a 2-regime SETAR model with a long-memory process in the first regime and a short-memory process in the second regime. We briefly introduce the properties of this model. Then, methods for locating the threshold parameter are proposed: we determine a plausible value for the threshold and estimate the other parameters of the model conditionally on this threshold value. Such a process is applied to stock indices and individual asset prices. A comparison with simple FARIMA models is made using some forecastibility criteria. Our empirical results suggest that our model offers an interesting alternative competing framework to describe the persistent dynamics in modeling the first two even powers of the returns.