Economic Links and Cross-Predictability of Stock Returns: Evidence from Characteristic-Based “Styles
Review of Finance, Volume 23, Issue 2, March 2019, Pages 363–395,
News released by one firm may contain value-relevant information also for other, economically-related firms. Such information spillovers are most obvious in the context of industry affiliations. For instance, it is reasonable to expect that Nike’s quarterly earnings also provide valuable information about the prospects of Adidas. While the efficient market hypothesis predicts that such value-relevant information is instantaneously reflected in security prices, prior research shows that prices often appear to react with a delay. This gives rise to a cross-predictability effect in returns from one stock to another.
I extend the literature on economic links and return predictability by focusing on a further channel of information transfers which is not captured by industry membership. Specifically, I investigate information spillovers among firms that share similar stock characteristics (such as having a high book-to-market equity ratio) and therefore can be classified as “similar-style” stocks (such as being a value stock). Coming back to the above example, Adidas and Nike do not only operate in the same industry, but they also share many other firm characteristics such as high past growth, high-profit margins, and relatively low book-to-market equity ratios.
Several considerations support the idea that characteristics capture economic links between firms. For instance, the fact that the returns of similar-style stocks tend to co-move suggests that common factors may also explain their earnings. Moreover, characteristic-based long-short strategies often generate abnormal returns with regard to standard asset pricing models. Such anomaly returns may compensate for exposure to a systematic, yet unknown risk factor, and therefore may reflect a common, systematic cash flow risk among similar-style firms.
To test the style-based information transfer hypothesis, I synthesize information from a total of 17 different firm-specific characteristics from the anomalies literature to classify stocks into different styles. The information set consists of earnings surprises, i.e. abnormal three-day returns, around quarterly announcement dates. Since earnings surprises proxy for unexpected news about current and future business conditions, they are arguably highly relevant for economically linked firms.
My main empirical analyses focus on the predictability of style returns and individual stock returns. I find that past style-based earnings surprises have statistically significant predictive power for 10 of the 17 style returns. For individual stocks, I develop a composite Style-based Earnings Surprise Measure (“SESM”). In multivariate firm-level regressions, SESM is a highly significant predictor of future stock returns, even among the largest firms in the sample with a market value above the NYSE median. Portfolio tests show that an equal-weighted (value-weighted) long-short strategy based on SESM realizes an unconditional return of 1.67% (1.01%) per month.
I do not find that industry spillovers, the traditional post-earnings announcement drift, unconditional abnormal style returns, or risk can explain this return predictability. In addition, I provide a variety of further results which are consistent with a gradual diffusion of fundamental information at the style level that causes this cross-predictability effect.