Time-Varying Factor-Augmented Models for Volatility Forecasting

2025-08-03

AI Summary

Existing volatility forecasting methods suffer from two key limitations: multivariate models lack computational scalability, while factor models rely on static loadings, failing to capture the time-varying nature of volatility co-movement. This paper proposes a general factor-augmented forecasting framework whose core innovation is the incorporation of time-varying factor loadings. Leveraging a dynamic factor model, it extracts low-dimensional common volatility factors that evolve jointly across both cross-sections and time, and seamlessly integrates them into statistical and AI-based models—including linear regression and machine learning—to jointly model asset-specific and market-wide volatility dynamics. The framework enables efficient multi-step forecasting for large-scale portfolios, offering both flexibility and scalability. Empirical results demonstrate substantial improvements in 1-day and 7-day volatility forecasts for U.S. technology stocks and major cryptocurrencies. Economic evaluation further shows that pairing-trading strategies built upon these forecasts yield superior risk-adjusted returns and profit stability, even under extreme market conditions.

Technology Category

• Machine Learning: Probabilistic Circuits and Graphical Models
• Game Theory and Economic Paradigms: Adversarial Learning
• Multiagent Systems: Adversarial Agents

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Accurate volatility forecasts are vital in modern finance for risk management, portfolio allocation, and strategic decision-making. However, existing methods face key limitations. Fully multivariate models, while comprehensive, are computationally infeasible for realistic portfolios. Factor models, though efficient, primarily use static factor loadings, failing to capture evolving volatility co-movements when they are most critical. To address these limitations, we propose a novel, model-agnostic Factor-Augmented Volatility Forecast framework. Our approach employs a time-varying factor model to extract a compact set of dynamic, cross-sectional factors from realized volatilities with minimal computational cost. These factors are then integrated into both statistical and AI-based forecasting models, enabling a unified system that jointly models asset-specific dynamics and evolving market-wide co-movements. Our framework demonstrates strong performance across two prominent asset classes-large-cap U.S. technology equities and major cryptocurrencies-over both short-term (1-day) and medium-term (7-day) horizons. Using a suite of linear and non-linear AI-driven models, we consistently observe substantial improvements in predictive accuracy and economic value. Notably, a practical pairs-trading strategy built on our forecasts delivers superior risk-adjusted returns and profitability, particularly under adverse market conditions.

Problem

• Existing volatility forecasting methods lack computational feasibility for realistic portfolios.
• Static factor models fail to capture evolving volatility co-movements effectively.
• Current approaches struggle with joint modeling of asset-specific and market-wide dynamics.

Innovation

• Time-varying factor model for dynamic volatility
• Integrates factors into statistical and AI models
• Improves predictive accuracy and economic value