Advanced Research in Economics and Business Strategy Journal
Volume 7, Numéro 1, Pages 219-238
2026-06-30

Geopolitical Risk And State-dependent Crude Oil Volatility Forecasting: Evidence From Rolling Out-of-sample Prediction

Authors : Juan Zhang . Wei Chong Choo . Choy Leong Yee . Yihuan Lin . Zhou Xuan .

Abstract

This study examines whether geopolitical risk contains incremental predictive information for crude oil market volatility in a rolling out-of-sample forecasting framework. Rather than focusing only on in-sample associations, the analysis evaluates whether the geopolitical risk index improves real-time prediction relative to a benchmark autoregressive model, denoted AR p. The study begins with return forecasting, where the evidence is limited and does not support a stable full-sample predictive role for geopolitical risk. Motivated by this result, the analysis shifts to volatility forecasting and considers two alternative proxies, namely absolute returns and squared returns. These measures capture different aspects of market risk, as absolute returns are more robust to extreme observations, whereas squared returns place greater weight on tail events. The empirical design compares a benchmark ARp\ model with an augmented ARp+GPR specification under both full-sample and regime-dependent evaluation. High-risk periods are identified from the upper tail of the geopolitical risk distribution. The results show that geopolitical risk does not generate uniform forecasting gains across the full sample. However, its predictive contribution becomes more relevant during high-risk periods, where state-dependent improvements emerge in volatility forecasting, although the strength and form of these gains vary across volatility proxies. Overall, geopolitical risk is better interpreted as a stress-state risk indicator than as a general-purpose predictor of crude oil market movements.

Keywords

Geopolitical Risk ; Crude Oil Volatility ; Out-Of-Sample Forecasting ; State Dependence ; Volatility Proxies