Statistical Analysis of Realized Volatility of Bitcoin Price using Heterogeneous Autoregressive and Generalized Autoregressive Conditional Heteroskedasticity Models

Bitcoin has recently gained extra attention in the financial industry and the blockchain community in general; it’s considered the most popular form of technology. As a result, the purpose of the study is to predict the actual volatility of the bitcoin price using generalized autoregressive conditional heteroskedasticity (GARCH) and heterogeneous autoregressive (HAR). In this research, the researcher attempted to utilize the appropriate statistical methodology, such as GARCH and HAR models. GARCH models were created to address the issue of volatility aggregation, which is the tendency for prices to cluster together as large changes occur. With the GARCH model, we can represent the conditional heteroskedasticity and the fat tail of financial market data. The primary objective was to directly observe and predict the behavior of volatility in time series data. Overall, the model’s architecture appears simple and is capable of reproducing the primary characteristics of financial information. The primary concept of this model is that investors with different time frames perceive and respond to different levels of volatility. Sample information about the price of the bitcoin cryptocurrency was distributed worldwide. It includes daily updates of the variable for the time period 31-Jun-17 to 31-Jan-22. The investigation has demonstrated that the HAR model is more effective at predicting variance for this period in comparison to GARCH (1, 1). The result shows that 1 day of previous year’s variance estimates and jump estimates have a significant impact on the future variance (h = 1).