Forecast value-at-risk for the cryptocurrency market using Markov-switching EGARCH models

Abstract

This study aims to understand the volatile behavior of six highly representative cryptocurrencies. To do so, EGARCH and Markov-switching EGARCH models were estimated, combined with different distributions of statistical probability. The predictive capacity of the best models resulting from these combinations were tested by predicting the value-at-risk. The daily returns of the cryptocurrencies clearly show regime changes in their volatility dynamics. In the in-sample analysis, the regime change model confirms the existence of two states: the first characterized by a greater ARCH effect and less affected by asymmetries, while the second reveals a greater effect of the arrival of information, that is, it is more sensitive to asymmetric shocks. In the out-of-sample analysis, the value-at-risk predictions of the regime change model clearly exceed the single-regime model by the extreme quantile of 1%.