Early Warning of Impending Oil Crises Using the Predictive Power of Online News Stories

Extreme events (such as natural disasters, political upheaval, economic crises) typically have a strong impact on crude oil markets and related price fluctuations and may eventually emerge to global oil crises. This study attempts to early detect such events based on the predictive power of online news messages. Text mining algorithms are used to turn unstructured news into actionable information and to determine which news can be regarded as relevant for the oil market. Over 45 million news messages have been examined. A decision support system is constructed which uses an indicator metric to set off an alarm based on information gathered from current and historic news stories. Regression analyses statistically attest the predictive power of online news messages and thus demonstrate the potential of the early warning system. The effect on the price of crude oil is statistically significant.