Investor Sentiment and (Anti) Herding in the Currency Market: Evidence from Twitter Feed Data

Abstract

This paper establishes a direct link between (anti) herding behavior in currency markets and investor sentiment, proxied by a social media based investor happiness index built on Twitter feed data. Our analysis of daily data for nine developed market currencies suggests that the foreign exchange market is generally characterized by strong anti-herding behavior. Utilizing the quantile-on-quantile (QQ) approach, developed by Sim and Zhou (2015), we show that the relationship between investor sentiment and anti-herding is in fact regime specific, with anti-herding behavior particularly prominent during states of extreme investor sentiment. The effect of sentiment on anti-herding is generally stronger in extreme bullish sentiment states, while average sentiment is associated with less severe anti-herding. The findings lend support to the behavioral factors for asset pricing models and suggest that real time investor sentiment signals can be utilized to monitor potential speculative activities in the currency market.

Keywords:

• Herding
• Exchange rates
• Time-varying regression
• Investor happiness

Notes

1 https://www.globalfinancialdata.com

2 https://www.bis.org/statistics/eer.htm

3 The daily trade weights are shown in Figure B.1 in Appendix B.

4 The happiness index is available at https://hedonometer.org/timeseries/en_all/

5 Using the Huber (1967) and White (1982) estimator.

6 Using the Roll Eviews add-in found at https://www.eviews.com/Addins/addins.shtml.

7 Other non-linear approaches to herding include Babalos and Stavroyiannis (2015b), Babalos et al. (2015)), and quantile regressions (e.g. Klein 201388), among others.