Uncertainty and oil volatility: Evidence from shrinkage method

Prior studies argue economic policy uncertainty index (EPU) can successfully predict oil price volatility. This paper mainly sheds light on the forecasting ability of categorical EPU indexes for both WTI and Brent oil futures. To illustrate this problem, we extend the LASSO model with Markov two-stage regimes (MS-LASSO model) to better fit the nonlinear feature of oil volatility dynamic. Our empirical results provide empirical evidence that EPU index not only contains predictive information content for WTI and Brent oil volatility, but achieves higher realized utility. Notably, we find that the predictive ability of categorical EPU indexes is asymmetric through global financial crisis, business cycle and different market conditions. Additionally, the uncertainty index of sovereign debt and currency crises is the most frequent predictor selected by shrinkage method for both WTI and Brent oil volatility.