Since the development of the capital asset pricing model (CAPM) in the 1960s, hundreds of anomalies (what John Cochrane famously called a “zoo of new factors”) have been uncovered in the cross-section of stock returns. While some of the anomalies (such as the size and value factors) have risk-based explanations, others (such as momentum) have behavioral-based explanations and thus demonstrate market inefficiencies (mispricings). Among the theories explaining the behavioral anomalies is that various biases (such as anchoring, loss aversion, overconfidence, endowment and cognitive dissonance) lead to investors’ inability to price news correctly. (For those interested, read an earlier article in Alpha Architect that covers investors’ misreaction to news.)
Ying Wang, Bohui Zhang and Xiaoneng Zhu contribute to the literature on asset pricing anomalies with their October 2018 study “The Momentum of News.” Relying on a comprehensive data set of news releases, they constructed monthly firm-level news sentiment scores over the 2000-2016 period. Specifically, RavenPack News Analytics (a provider of big data analytics) quantifies the positive (or negative) information (i.e., news sentiment score) in each news article based on professional algorithms. For example, a news article on a corruption scandal involving a firm’s executives is associated with a low news sentiment score, and a news article regarding the successful development of a firm’s new product is associated with a high news sentiment score. Their final sample consisted of 530,283 firm-month observations (2,600 firms on average).
The following is a summary of their findings:
- There is a news momentum phenomenon. Stocks with more positive news in the past generate more positive news in the future, with the news momentum persisting for two years.
- There is a monotonic relationship between current and future news sentiment scores—firms with more current positive (negative) news are likely to release more positive (negative) news in the future.
- News momentum is driven by the persistence of firms’ fundamentals rather than stale news or firms’ strategic disclosure of information.
The News Momentum Strategy
A monthly trading strategy that combines a long position in a good news quintile portfolio with a short position in a bad news quintile portfolio generates a monthly alpha of 0.736 percent risk-adjusted (risk-adjusted via the Fama-French three factors), which is significant at the 1 percent confidence level. The five-factor alpha was 0.716 percent per month and was also statistically significant at the 1 percent confidence level. However, at longer horizons (one year or longer), the profits become economically insignificant (indicating the premium is not explained by risk). It is also significant only for stocks with poor information environments, such as those with small firm size, low analyst coverage and fewer institutional holdings, consistent with the mispricing view of return predictability—as small firms attract less attention and have fewer news releases, information is likely to be more asymmetric and to be diffused more slowly for these stocks.
- The return anomaly appears on both news and non-news days.
- News-driven return predictability is caused by investors’ underreaction to both news momentum and news itself.
- The results hold for various tests of robustness.
Wang, Zhang and Zhu concluded that their findings “suggest that the cross-sectional prediction of news is not fully incorporated into the stock price by investors.”
It’s important to note that their results do not include any estimates of implementation costs, and no estimates of turnover are provided. That said, the alpha seems sufficiently high that it is likely it would survive trading costs, especially since the fact that news momentum persists should mean that turnover should not be excessively high—both the positive and negative news scores tend to persist for up to two years and longer. Finally, the fact that the premium was found to be significant only for small stocks with low analyst coverage and fewer institutional holdings provides an argument for limits to arbitrage, allowing the anomaly to persist.
The findings of Wang, Zhang and Zhu are consistent with that of prior research, which found that momentum is really fundamental momentum. For example, in his February 2015 National Bureau of Economic Research paper, “Fundamentally, Momentum is Fundamental Momentum,” Robert Novy-Marx demonstrated that momentum in stock prices is driven by fundamental momentum—it’s “a weak expression of earnings momentum, reflecting the tendency of stocks that have recently announced strong earnings to outperform, going forward, stocks that have recently announced weak earnings.” (Read a summary of Novy-Marx’ paper in Alpha Architect, and a deeper dive on the concept here.)
Marx’s findings are consistent with those of Shuoyuan He and Ganapathi Narayanamoorthy, authors of the October 2017 study, “Earnings Acceleration and Stock Returns.” They found that earnings acceleration is a significant predictor of future stock returns. The authors concluded that the future return predictability appears to arise because of investors missing predictable implications of earnings acceleration for earnings growth two and three quarters hence—investors appear to underestimate the magnitude of the effect of earnings acceleration on two- and three-quarters-ahead earnings growth. He and Narayanamoorthy also concluded that the evidence of the extremely high persistence of a positive return to the strategy indicates that there is not a likely risk-based explanation for the high returns. Thus, the returns are likely a result of a behavioral anomaly. Which begs the question of why investors are so myopic and do not price the implications of earnings acceleration two and three quarters hence. What are the behavioral underpinnings of such a bias?
Wang, Zhang and Zhu provide new insights into the momentum factor, helping us understand the source of returns to this anomaly. Their findings are consistent with those of prior research, providing a good example of how behavioral anomalies can persist post-publication when there are sufficient limits to arbitrage present, preventing sophisticated investors from correcting mispricings.