Threshold effects in oil–metal volatility spillovers: Evidence from industrial and precious metals

Understanding when and how crude oil volatility transmits to metal markets is critical for managing risk in volatile commodity environments. Although oil–metal linkages have been widely studied, it remains unclear whether volatility spillovers depend on the intensity of oil-market turbulence, whether fixed or time-varying thresholds better characterize these dynamics, and whether industrial and precious metals respond differently. To address these issues, this study examines regime-dependent volatility spillovers from crude oil to industrial and precious metals. Using threshold autoregressive (TAR) and conditional threshold autoregressive (CoTAR) models, the analysis assesses whether metal volatility responds differently when crude oil volatility crosses fixed or time-varying thresholds. It further distinguishes total, good (upside), and bad (downside) realized volatility components to evaluate whether risk transmission is symmetric across volatility states or concentrated in adverse market conditions. In-sample results reveal threshold effects for industrial metals, particularly tin, nickel, and zinc, where spillovers from crude oil volatility intensify under extreme market conditions, especially during downside-volatility episodes. Among precious metals, threshold effects are less systematic, with gold and silver showing limited evidence of statistically significant regime shifts. Out-of-sample results support these asymmetries, showing that threshold models improve predictive accuracy for industrial metals, especially during downside-risk episodes, whereas a linear autoregressive model performs best for precious metals. These findings highlight the state-dependent and metal-specific nature of oil–metal volatility transmission, showing that spillovers should not be treated as uniform across metals or volatility components. They also suggest that incorporating oil-volatility regimes may improve volatility monitoring and forecast model selection for selected industrial metals, with more limited gains for precious metals.