Regime Shift Model by Three Types of Distribution Considering a Heavy Tail and Dependence

Abstract

I adopt a regime shift model to investigate a shift of distribution of each regime during a time series data. Unlike previous studies, I applied three types of distribution to use a regime shift model, i.e., normal, GEV and stable distribution, which allows me to consider a heavy tail regime in the model. From some theoretical basis and empirical results, I find that the regime shift model in stable distribution is best appropriate. I also find that tail index of the innovation and dependence measure move together, implying dependence among a consecutive data may lead extreme event and vice versa.