Symmetry or Asymmetry: A model comparison between different ARCH-class volatility models using Bitcoin returns

Sammanfattning: This thesis will in turn evaluate the forecast performance of different ARCH-type models' forecast ability using Bitcoin returns from 01-04-2015 to 01-04-2022. More specifically, it is of interest to see if a simple GARCH(1,1) model can outperform more sophisticated models that incorporate the asymmetry in volatility. Besides the GARCH(1,1) model, other models used in this thesis were the IGARCH, EGARCH, GJR-GARCH, TGARCH and APARCH models, where the EGARCH, GJR-GARCH, TGARCH and APARCH are asymmetric volatility models. To compare the different models' forecast abilities, the Hansen et al. (2011) Model Confidence Set test was utilized. The results showed that the case for using asymmetric models to explain Bitcoin returns are limited, as the GARCH(1,1) were often selected by the Model confidence set. Although the overall performances of asymmetric models were poor at all forecast horizons, the GJR-GARCH model performed quite admirably, as it was selected more times by the model confidence set test than the GARCH(1,1). On the other hand, even though the GJR-GARCH performed better than the GARCH(1,1), this is not sufficient evidence to warrant the use of asymmetric models, as the other asymmetric models were, more often than not, selected by the model confidence set procedure