Title: Detecting market irrationality using news sentiment and information entropy
Authors: Jing Chen - Cardiff University (United Kingdom)
Anqi Liu - Cardiff University (United Kingdom) [presenting]
Steve Yang - Stevens Institute of Technology (United States)
Abstract: News sentiment, an investor sentiment proxy, has been widely explored in behavioral finance; yet the linkage between investor sentiment and market irrationality and inefficiency has not been thoroughly examined. We consider the financial market as a bivariate system that consists of news sentiment and market returns. We adopt the concept of transfer entropy to quantify information flow between these two types events and formulate irrationality regime proxies. Testing with an intraday dataset from 2003 to 2014 for the major U.S. markets, we find that the information flow follows a trimodal distribution that clearly distinguishes financial markets into three regimes: the price-driven, transitional and news-driven regimes. We provide evidence to show that the proposed irrationality proxy is positively correlated with three market inefficiency indicators in the current literature; and also we identify a significant cut-off threshold to delineate the market into price-driven and news-driven regimes, showing that news-driven investment decision is a key factor of market inefficiency.