Supply disruptions and predictability of oil return volatility: a GARCH-MIDAS approach

Abstract We utilize a Generalized Autoregressive Conditional Heteroskedasticity-Mixed Data Sampling (GARCH-MIDAS) framework to show that monthly indicators of global and United States supply concerns, along with their regional- and product-specific components, constructed from 200 million shipment-level transactions, increase daily West Texas Intermediate (WTI) oil returns volatility over the period from January 2013 through August 2024. Furthermore, the in-sample predictability of oil returns volatility translates into statistically significant and robust short-, medium-, and long-term forecasting gains derived from the information content of the various supply disruption indexes, relative to a benchmark model of GARCH-MIDAS-realized volatility (RV), over the out-of-sample period of January 2020 and beyond. The comprehensive supply disruption indexes are also shown to outperform other widely used monthly measures of global supply constraints and economic conditions when incorporated into the GARCH-MIDAS framework. Our results carry important implications for investors and policymakers.