Journal of Advanced Robotics, Autonomous Systems and Human-Machine Interaction
Modeling and Forecasting Bitcoin Volatility: A Comparative Analysis of GARCH and GJR-GARCH Models
Abstract
Waseem Khoso and Muhammad Rafique Daudpoto
This study conducts a comparative analysis of GARCH-family models to capture the volatility dynamics of Bitcoin. Utilizing a comprehensive dataset of daily Bitcoin prices, we find that the return series exhibits classic financial stylized facts, including stationarity, leptokurtosis, and negative skewness. Through rigorous model selection criteria and diagnostic testing, the GARCH (1,3) model demonstrates a marginally superior fit compared to the asymmetric GJR-GARCH (1,1) model. Intriguingly, the leverage effect parameter in the GJR-GARCH model is statistically insignificant, suggesting that during our sample period, Bitcoin’s volatility response to positive and negative news shocks was largely symmetric. Our findings provide critical insights for risk managers and financial analysts modeling cryptocurrency volatility.