Impact of media hype and fake news on commodity futures prices: A deep learning approach over the COVID-19 period

Abstract

We investigate the reactions of eight commodity futures to media hype and fake news during COVID-19, utilising the Ravenpack news database, along with deep learning algorithms. Results identify a significant impact on commodity prices of media hype and fake news, with this reaction amplified during COVID-19. Compared to alternative deep learning algorithms, bi-directional long-short-term memory is adaptive to forecasting the returns of the commodity futures contracts with lower mean absolute error and root mean square error. Findings, confirmed by Diebold-Mariano testing, as well as alternative data partitioning, show commodity markets are susceptible to fake news and media hype.