FORECASTING BITCOIN VOLATILITY USING A WEIGHTED SEMIPARAMETRIC EGARCH MODEL INCORPORATING CRUDE OIL MARKET VOLATILITY

Abstract

Bitcoin price movements are highly volatile and nonlinear over time. Traditional GARCH models are widely used for volatility modeling, but they may not adequately represent nonlinear effects of exogenous variables. In this study, we develop a weighted semiparametric EGARCH model for forecasting Bitcoin volatility by combining the parametric conditional variance from an EGARCH model with a nonlinear crude oil market volatility component estimated using the Nadaraya–Watson kernel estimator. The empirical analysis uses daily Bitcoin and crude oil price data from 2017 to 2026. The forecasting performance is compared with GARCH(1,1), EGARCH(1,1), and semiparametric EGARCH models incorporating crude oil market volatility. The root mean square error, mean absolute error, quasi-likelihood loss, and Diebold–Mariano test are computed to assess the out-of-sample forecast accuracy for various training-testing split designs. Empirical results indicate that the proposed weighted semiparametric EGARCH model achieves higher forecasting accuracy and robustness for forecasting Bitcoin volatility.