The predictive power of public Twitter sentiment for forecasting cryptocurrency prices

Olivier Kraaijeveld, Johannes De Smedt

Research output: Contribution to journalArticlepeer-review


Cryptocurrencies have become a very popular topic recently, primarily due to their disruptive potential and reports of unprecedented returns. In addition,academics increasingly acknowledge the predictive power of Twitter for a wide variety of events and more specifically for financial markets. This paper studies to what extent public Twitter sentiment can be used to predict price returnsfor the nine largest cryptocurrencies: Bitcoin, Ethereum, XRP, Bitcoin Cash,EOS, Litecoin, Cardano, Stellar and TRON. By using a cryptocurrency-specific lexicon-based sentiment analysis approach, financial data and bilateral Granger causality testing, it was found that Twitter sentiment has predictive power for the returns of Bitcoin, Bitcoin Cash and Litecoin. Using a bullishness ratio,predictive power is found for EOS and TRON. Finally, a heuristic approach is developed to discover that at least 1-14% of the obtained Tweets were posted by Twitter "bot" accounts. This paper is the first to cover the predictive power of Twitter sentiment in the setting of multiple cryptocurrencies and to explore the presence of cryptocurrency-related Twitter bots.
Original languageEnglish
JournalJournal of International Financial Markets, Institutions and Money
Early online date12 Mar 2020
Publication statusE-pub ahead of print - 12 Mar 2020


  • cryptocurrencies
  • time series analysis
  • sentiment analysis
  • Natural Language Processing
  • Twitter
  • bots

Fingerprint Dive into the research topics of 'The predictive power of public Twitter sentiment for forecasting cryptocurrency prices'. Together they form a unique fingerprint.

Cite this