Modeling Volatility for High-Frequency Data of Cryptocurrency Bitcoin Price using Generalized Autoregressive Conditional Heteroskedasticity (GARCH) Model

Nashirah Abu Bakar, Sofian Rosbi

Cryptocurrency, Bitcoin, Volatility, GARCH Model, High-Frequency Data.

The cryptocurrency namely Bitcoin is a decentralized cryptocurrency considered a type of digital asset that uses public-key cryptography to record, sign and send transactions over the Bitcoin blockchain. All transaction processes are performed without the oversight of a central authority. The time series data for Bitcoin price movement exhibit time-varying volatility and volatility clustering. This study aims to evaluate the time-varying volatility of Bitcoin price using the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model. This study uses daily share prices starting from July 2017 until July 2022. The mean equation was developed using the ARMA (1,1) for Bitcoin return. Next, this study evaluated OLS, GARCH, GARCH-M, and E-GARCH models. The result shows the EGARCH (1,1) model exhibits its lowest error of AIC with a value of 5.5984. The autocorrelation test was performed using Q-statistics indicating EGARCH (1,1) model is free from the autocorrelation problem. In addition, ARCH-LM test indicates EGARCH (1,1) is free from heteroscedasticity problems. The EGARCH (1,1) shows there is a leverage effect for volatility clustering. This explained the behavior of bad news effect more than positive news. The finding of the study can act as a guideline to help investors to analyze their investment behavior. At the same time, the finding of this study helps investors to understand the cryptocurrency dynamics behavior.

Received: 28 Aug 2022, Received in revised form: 20 Sep 2022, Accepted: 24 Sep 2022, Available online: 30 Sep 2022

10.22161/ijaers.99.62