<![CDATA[This study examines the prediction ability of Global Financial Stress Index (GFSI) and Geopolitical Risk Index (GPR) to forecast the volatility of assumed safe-haven assets, like gold, silver and Bitcoin. The authors combine high-frequency model, Heterogeneous Autoregression Realized Volatility (HAR-RV) model, with low-frequency estimators that use low-frequency data, such as Parkinson, Garman-Klass, and Rogers-Satchel volatility estimators. The authors also examine the R2 out-of-sample of the created model to conclude that GFSI and GPR can be used to increase the forecasting accuracy of selected asset's volatility and demonstrate the effectiveness of variables to be used as predictive variables. According to the result, the R2 out-of-sample of models that used GFSI as predictive variable have better performance in forecasting on gold, silver and Bitcoin. Meanwhile, GPR is assumed not effective as GFSI to be a predictive variable. The RB-HAR-ASY-GFSI model can increase the forecasting accuracy up to 1.56% (5-day ahead prediction) on gold, up to 0.43% (66-day ahead prediction) on silver, and up to 2.78% (10-day ahead prediction) on Bitcoin. This study improves the undersWangding of financial and geopolitical uncertainty impact on the volatility of safe-haven assets. Second, it investigates the HAR-RV and low-frequency data combination model performance to forecast volatility assets.]]>
Copyrights © 2025
