Abstract
This paper develops a set of innovative text-based indices capturing oil market sentiment and oil price uncertainty. The textual analysis includes over 6 million news items spanning from January 1982 to June 2021. The evidence shows that sentiment indicators readily react to economic and geopolitical events impacting oil prices, thereby enabling said indicators to accurately predict the price of oil. In contrast, uncertainty measures have inherent weaknesses thus yielding unreliable oil price forecasts. This research results in a novel and robust text indicator that offers valuable insights for predicting the intricate dynamics of crude oil prices, particularly excelling in short-term forecasts and during periods of economic recessions.
| Original language | English |
|---|---|
| Volume | Summer 2025 |
| Specialist publication | Commodity Insights Digest |
| Publication status | Published - Mar 2025 |
Keywords
- text mining
- Bayesian vector autoregression
- stochastic volatility
- density forecast
- ROC