A0856
Title: Regime dependent jump frequencies in cryptocurrency log returns
Authors: Mai Phan - University of Kaiserslautern-Landau & HTW Berlin (Germany) [presenting]
Joern Sass - RPTU Kaiserslautern-Landau (Germany)
Christina Erlwein-Sayer - University of Applied Sciences HTW Berlin (Germany)
Abstract: The cryptocurrency market has risen globally over time. The market is characterized by strong volatility. Cryptocurrencies series are heteroscedastic, not normally distributed and show volatility clustering, which indicates strong price fluctuations. Cryptocurrencies' prices have been researched to analyse their characteristics and minimize the risk. Traditional time series methods fail to capture their extreme fluctuations. Other studies have shown that hidden Markov models are a good choice to model price predictions. They cover sudden movements from the market by considering different states. Nevertheless, those models do not consider jump occurrences of log returns, although those frequencies change over time. The idea is to model log returns of cryptocurrencies with a regime-switching GARCH-jump model. This model includes not only the regime-specific jump frequencies as well as various states but also GARCH processes depending on regimes for the conditional variance. With these specifications, more flexibility is added. The proposed approach is used to model the daily log returns of Bitcoin.
