Time-varying predictability of oil market movements over a century of data: The role of US financial stress

In this paper we analyze whether a news-based measure of financial stress index (FSI) in the US can predict West Texas Intermediate oil returns and (realized) volatility over the monthly period of 1889:01 to 2016:12, using a dynamic conditional correlation multivariate generalized auto-regressive conditional heteroscedasticity (DCC-MGARCH) model. Our results show that, standard linear Granger causality test fail to detect any evidence of predictability. However, the linear model is found to be misspecified due to structural breaks and nonlinearity, and hence, the result of no causality from FSI to oil returns and volatility cannot be considered reliable. When we use the DCC-MGARCH model, which is robust to such misspecifications, in 75 percent and 80 percent of the sample periods, FSI in fact do strongly predict the oil returns and volatility respectively. Overall, our results highlight that FSI is helpful in predicting oil returns and volatility, when one accounts for nonlinearity and regime changes through a robust time-varying model.