Multi-frequency GARCH modelling for high-frequency volatility forecasting in foreign exchange markets

Read full text Summarise with VIRAT
Authors: Aidan Van Klaveren
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study investigates how sampling frequency affects the performance and internal coherence of GARCH-based volatility forecasting models in high-frequency foreign exchange markets. Using minute-level data across multiple intervals (1-, 5-, 15-, 30-minute, and 1-hour), several specifications—including GARCH(1,1), EGARCH, TGARCH, and GJR–GARCH with Student-t innovations—were evaluated within a rolling-window, walk-forward validation framework designed to prevent information leakage. Forecast accuracy was assessed using MSE, RMSE, and MAE, complemented by the Diebold–Mariano test to evaluate statistical differences in predictive performance across time scales.In addition to direct forecasts, higher-frequency variance predictions were aggregated to approximate lower-frequency realised volatilities, enabling cross-scale comparison of model coherence. Results show that aggregated forecasts frequently match or outperform direct lower-frequency predictions, demonstrating that fine-scale information enhances volatility estimation even after temporal aggregation. A triangular exchange-rate diagnostic based on the GBPEUR– USDEUR–USDGBP system further examined whether independently estimated univariate GJR–GARCH models maintain internal consistency under the standard no-arbitrage assumption. Mean Absolute Relative Error (MARE) and correlation-gap metrics confirmed that the modular framework produces stable and coherent forecasts across currencies and frequencies.Overall, the findings demonstrate that modular, high-frequency GARCH models can achieve both empirical accuracy and internal coherence across time horizons. This supports their use as scalable, adaptive components in real-time volatility forecasting and cross-currency risk management systems.

Fields of Study & Subjects

Financial Risk and Volatility Modeling Stock Market Forecasting Methods Stochastic processes and financial applications

Publication details

Journal / source
Macquarie University
Publication date
30 March 2026
Language
EN
Document type
Thesis