This study investigates the transitory and persistent intraday volatility linkages among the U.S. futures markets for oil, gold, equities, and foreign exchange. The main objective is to disentangle and analyze short-term and long-term volatility connectedness across these markets using the novel frequency connectedness approach of Baruník and Ellington (2024). A secondary objective is to evaluate the predictive power of macroeconomic, financial, and geopolitical indicators using machine learning models. By integrating high frequency data with advanced forecasting techniques, the study offers new insights into the structure, drivers, and forecastability of volatility spillovers across time horizons. The findings reveal that volatility connectedness is significantly more pronounced in the short-term. FX and gold emerge as dominant transmitters of volatility across all time horizons. In contrast, the oil market is more vulnerable to external shocks in the short-term, rather than transmitting volatility. Among the machine learning models, the Decision Tree Forest achieves the highest forecasting accuracy for both transitory and persistent connections. Moreover, the analysis identifies VIX, GPR, and the ADS Index as key drivers of short-term and long-term connectedness dynamics. These results provide valuable insights for policymakers, investors, and portfolio managers, emphasizing the interconnected nature of financial markets and the role of uncertainty in shaping volatility.
