| The primary determinant of equity portfolio risk is the likelihood that pairs of stock prices will rise and fall together. Both in research and in practice, our expectations regarding future price comovement, and our appraisals of portfolio risk, have relied on lengthy series of historical stock returns. Yet, the mechanism generating these returns depends on a flow of firm-specific information that changes throughout time. As new sources of opportunity and uncertainty are revealed to the market, the links between distant historical prices and future stock price comovement become weaker. Therefore, accurate predictions of comovement must also consider the similarity of contemporaneous information flows across firms. I develop a proxy for this similarity and test whether this new measure can improve predictions of future stock price comovement. The field of finance is replete with simple quantitative descriptors designed to identify similarities in firm characteristics. However, contemporaneous changes in the flow of information are not reflected in these quantitative measures until firms announce earnings or publish financial reports. To identify contemporaneous changes in firm similarity, investors must rely on softer, more qualitative, sources of information. In real-time, this content is often delivered to the market through financial newswires. These services act as information conduits by compressing a vast array of firm-specific material into a digestible sequence that investors can use to make portfolio decisions. This paper examines whether the qualitative information circulated on one such newswire, the Reuters Integrated Data Network, can predict how future equity payoffs are correlated across firms. During each six-month period from 2003 to 2013, I measure the similarity of firm-specific newswire text written about different companies. I propose that the contemporaneous information flows for two firms are qualitatively similar if there is commonality in their newswire text. In support of this hypothesis, I find that the qualitative similarity of the newswire items written about a firm pair predicts their stock return correlation during the following six-month period. Furthermore, this new measure of qualitative similarity can predict future price comovement even after accounting for the pair’s contemporaneous return correlation. Thus, qualitative similarity describes similarity in information flows that cannot be inferred from historical stock prices. Prior literature has recognized that characterisitics such as firm beta (Ledoit and Wolf 2003), size (Pindyck and Rotemberg 1993), book-to-market (Bekaert et al., 2009), momentum (Asness et al., 2013) and industry (Campbell et al., 2001; Irvine and Pontiff, 2009; Brandt et al., 2010) proxy for common sources of systematic variance that generate price comovement between firms. The literature also offers many alternative explanations for stock price comovement that are based on some type of market friction. Specifically, cross-sectional variation in information diffusion (Barberis et al., 2005), as well as the categorical trading of assets (Barberis and Schleifer 2003), have been shown to cause higher levels of stock price comovement. To ensure that qualitative similarity does not proxy for one of these other documented sources of return correlation, I show that my measure’s predictability remains after controlling for similarities in exposure to systematic risk as well as firm liquidity, price, index membership, text volume and headquarters location. Thus, commonality in the information flow across firms predicts return correlation that cannot be accounted for with standard asset pricing models and alternative explanations for stock price comovement . |
