News Content, Investor Misreaction, And Stock Return Predictability
Whitman School of Management, Syracuse University
University of Texas at Dallas – Department of Accounting & Information Management
Using a large dataset of news releases, we study instances of investors’ mistaken reaction, or misreaction, to news. We define misreaction as stock prices moving in the direction opposite to the news when it is released. We find that news tone predicts returns in the cross-section only upon the occurrence of misreaction. Stocks that are larger, more liquid, more visible, and more covered, by analysts or by the media, are less likely to exhibit misreaction. On the other hand, the ambiguity and complexity of news content, and variables that proxy for investor distraction, are all associated with more misreaction and greater predictability.
News Content, Investor Misreaction, And Stock Return Predictability – Introduction
There is extensive evidence that stock prices underreact or overreact on average to certain types of news. While underreaction or overreaction has been shown to lead to predictability in the crosssection of stock returns, it can be difficult to identify ex ante specific instances when stocks underreact or overreact to news. In this paper, we study investors’ mistaken reactions, or misreactions, to news. We define misreaction as stock prices moving in the direction opposite to the news when it is released. We characterize the determinants of misreaction and show that misreaction predicts returns in the cross-section.
We employ a comprehensive dataset of all news released on the Reuters data feed to identify instances of investor misreaction to news and investigate the informational content of news releases in financial markets. Our main findings are easily summarized. First, we find that stocks with positive news content subsequently outperform those with negative news content. In poolpanel regressions that control for a battery of factors, we find that stocks that had positive news content during the prior month subsequently earn a monthly return 0.75% higher than stocks that had negative news content. This predictability applies mainly to smaller stocks: the difference in average monthly returns between high and low news tone stocks is 1.25% for stocks in the bottom 25% of stocks by market capitalization and 0.45% per month for other stocks.
Second, we partition the sample into two groups depending on whether the signs of news tone and stock returns match over the portfolio formation period. We view instances of the signs not matching as prima facie evidence of mistaken reaction on the part of investors, e.g., bad news that is accompanied by positive returns or vice versa. Consistent with this view, we find subsequent predictability only when stock prices move in the direction opposite to the news when the news is released. We thus find that misreaction to the information contained in news releases drives the predictability that we uncover.
Third, we characterize the frictions that are associated with misreaction and predictability. We consider several types of frictions. First, theories of limited stock market participation (e.g., Merton, 1987) suggest that the stock prices of less visible firms can experience delays in incorporating information. Also, theories of limited attention suggest that when the amount of attention investors direct towards a firm decreases, its stock price should exhibit more severe underreaction to news and greater predictability (Hirshleifer, Lim and Teoh, 2011). We find that proxies for investor recognition such as analyst coverage and institutional ownership are associated with less misreaction. We also find that the amount of news coverage about the firm itself (a proxy for investor attention to the firm) is associated with less misreaction, while news coverage about other firms (which directs attention away from the firm) is associated with greater misreaction. Similarly, we find more misreaction when the proportion of news stories about the firm that are released during non trading hours is greater.
See full PDF below.