Evaluating and Enhancing Value-at-Risk: A Regime-Switching Approach to U.S. Equity Risk

Abstract

This paper evaluates the performance of Value-at-Risk (VaR) models under different market volatility regimes and proposes a regime-switching (RS) approach to improve risk measurement in U.S. equity markets. Using daily data from Apple, Microsoft, and the S&P 500 index, the study compares three traditional VaR methods: Historical Simulation (HS), Variance–Covariance (VCV), and Monte Carlo (MC). Market regimes are identified through rolling realized volatility, where periods of unusually high volatility are classified as turbulent. The study applies rolling one-day-ahead VaR forecasts and evaluates model performance through violation rates, Kupiec coverage tests, and breach severity analysis. The findings indicate that traditional VaR models generally perform adequately during stable market conditions but tend to underestimate downside risks during turbulent periods, particularly during events such as the COVID-19 pandemic. The proposed regime-switching VaR model combines the efficiency of the VCV approach during stable periods with the adaptability of the HS approach during turbulent periods. Results demonstrate that the RS method provides more stable coverage, fewer VaR violations, and lower breach severity across different asset types. The study highlights the importance of adaptive and regime-aware risk management models in capturing changing market dynamics and improving financial risk assessment.