Journal of Banking & Finance Vol. 169 2024
Information spillover and cross-predictability of currency returns: An analysis via Machine Learning
Abstract
This paper documents significant cross-return predictability of news variables, derived from textual analysis of news articles, for a broad cross-section of currencies. By employing forecasts based on the Least Absolute Shrinkage and Selection Operator (LASSO) that incorporate both news variables and forward discounts, we develop a notably profitable trading strategy. This strategy proves robust against transaction costs, risk adjustments, and controls for currency characteristics. Further analyses indicate that both risks and market frictions contribute to the profitability of the trading strategy, highlighting the crucial role of news in financial markets.
- DOI
- 10.1016/j.jbankfin.2024.107313
- Volume
- 169
- Pages
- 107313
- Language
- en
- Sources
- bibtex:phds-export.bib openalex crossref