Historical Volatility Fluctuations of Bitcoin: Influenced by Real-World Event

Abstract

Bitcoin market has exhibited substantial volatility over time. Bitcoin returns exhibit high standard deviation. This study employs the GARCH (1,1) model with normal (norm), Studentt (std), and generalized error distributions (ged) to estimate Bitcoin conditional volatility. Bitcoin exhibits fat-tailed returns, volatility clustering, and a remarkably high persistence value. The GARCH (1,1)-ged model showed superior performance compared to other models when evaluated using LL, AIC, and BIC criteria. The indicator saturation (IS) method was employed to concurrently detect historical daily breaks, trend breaks, and outliers in Bitcoin volatility data. The indicator saturation approach revealed that, for the past decade, historical Bitcoin volatility has had 6 outliers, 31 breaks, and 74 trend breaks under the normal distribution, 0 outliers, 26 breaks, and 83 trend breaks under the student-t distribution, and 1 outlier, 29 breaks, and 77 trend breaks under the ged distribution. This shows that assuming a heavy tail led to fewer outliers and breaks, and as the frequency of trend breaks increases, it also shows more volatility clusters represented by GARCH. These discoveries have the potential to comprehend the influence of events on financial markets and guarantee stability in the evaluation of financial risk, management of portfolios, and modeling endeavors.