Predicting the volatility of major forex futures: The dynamic persistence model

The modeling and forecasting of volatility in foreign exchange (forex) futures represent a critical nexus of quantitative finance, risk management, and international macroeconomics. This research explores the evolution of volatility persistence in seven major forex futures markets—AUD, GBP, CAD, JY, NE, SF, and EUR—using the Time-Varying Extended Wold Decomposition (TV-EWD) framework. Traditional volatility models, such as GARCH and the static HAR framework, often assume fixed parameters, failing to capture the smoothly evolving heterogeneity in shock dynamics characteristic of modern markets. By integrating locally stationary processes, this study decomposes realized volatility into scale-specific components ranging from 2 to 128+ trading days. The methodological approach allows for the identification of "pockets of predictability" during periods of heightened global uncertainty, including the Eurozone debt crisis, Brexit, and the COVID-19 pandemic. Empirical results demonstrate that incorporating time-varying shock persistence systematically improves forecast accuracy, particularly at the monthly (22-day) horizon, yielding error reductions of up to 21.8 % over standard benchmarks. The study also validates model calibration through rigorous Value-at-Risk backtesting. These findings contribute a novel methodological bridge between static multiscale decomposition and dynamic parameter evolution, offering significant implications for derivative pricing, portfolio optimization, and strategic risk management in an interconnected and non-stationary financial environment.