Can joint modelling of external variables sampled at different frequencies enhance long-term Bitcoin volatility forecasts?

• DOI: 10.1016/j.frl.2024.106679
• Publication Date: 2024-12-20T19:15:08Z
• Authors: Serkan Aras; Mehmet Ozan Özdemir; Cihan Çılgın

While monthly and weekly indices are commonly used for long-term Bitcoin volatility modelling, this study examines the role of daily indices in forecasting. Additionally, we evaluate the incremental contribution of daily indices when combined with the more frequently employed monthly and weekly indices. The findings reveal that daily Economic Policy Uncertainty (EPU) and Geopolitical Risk (GPR) indices outperform their monthly counterparts in both in-sample explanatory power and out-of-sample forecast accuracy. Moreover, it has been observed that using indices at different frequencies together significantly improves predictive performance. This study, therefore, demonstrates that mixed-frequency indices offer complementary insights for modelling Bitcoin volatility.