Start Date
10-12-2017 12:00 AM
Description
Predictive indicators for financial markets based on online buzz has been a frequent topic during the last years. Recent studies use a range of alternative sources for building these sentiment indices, with each purporting to have predictive value. Therefore a question mark remains regarding the comparability of findings across different types of sources, e.g. do indicators based on Tweets perform equally well or better than those built on news? This study addresses how competing sentiment indicators affect EUR/USD trading. To identify the indicator having the best predictive value we estimate expected returns for individual sources and forecast models via backtesting. Our findings support the notion that the predictive value depends on the source of the sentiment-indicator, on timing aspects, with more recent sentiments having greater predictive strength, and on the type of rule (e.g. buy / sell) harnessed.
Recommended Citation
Janetzko, Dietmar; Krauss, Jonas; Nann, Stefan; and Schoder, Detlef, "Breakdown: Predictive Values of Tweets, Forums and News in EUR/USD Trading" (2017). ICIS 2017 Proceedings. 10.
https://aisel.aisnet.org/icis2017/DataScience/Presentations/10
Breakdown: Predictive Values of Tweets, Forums and News in EUR/USD Trading
Predictive indicators for financial markets based on online buzz has been a frequent topic during the last years. Recent studies use a range of alternative sources for building these sentiment indices, with each purporting to have predictive value. Therefore a question mark remains regarding the comparability of findings across different types of sources, e.g. do indicators based on Tweets perform equally well or better than those built on news? This study addresses how competing sentiment indicators affect EUR/USD trading. To identify the indicator having the best predictive value we estimate expected returns for individual sources and forecast models via backtesting. Our findings support the notion that the predictive value depends on the source of the sentiment-indicator, on timing aspects, with more recent sentiments having greater predictive strength, and on the type of rule (e.g. buy / sell) harnessed.