The momentum effect is stronger among firms with higher levels of newspaper media coverage. Excess profits attributed to excess media coverage are found among firms exhibiting higher levels of valuation uncertainty, among firms with a media tone consistent with their formation period return profile, and among firms headquartered in US states ranked as being more individualistic.
What’s Inside?
Investor overreaction is shown to be a possible explanation for the momentum effect, a return anomaly, based on studying the impact of newspaper media coverage on driving the momentum of stock returns. The authors consider level of coverage, article tone, long-run reversals, level of uncertainty, and level of investor individualism. They run broad robustness tests to control for and evaluate various considerations.
How Is This Research Useful to Practitioners?
The authors consider the predictability of returns by studying the return patterns for three stock categories: stocks in the lowest residual media coverage quintile, stocks in the highest residual media coverage quintile, and a subset of the stocks in the second category considering the tone of the articles. The predictability of initial excess returns may be higher for those firms with high media coverage, with a reversal observed beginning around 10 months after the portfolio construction. After 36 months, the cumulative buy-and-hold return difference approaches zero between the two long–short portfolios—one consisting of long positions in high-coverage winners and short positions in high-coverage losers and one consisting of long positions in low-coverage winners and short positions in low-coverage losers. These baseline findings suggest investor overreaction as an explanation for the momentum effect, which is consistent with the approach of Daniel, Hirshleifer, and Subrahmanyam (Journal of Finance 1998), who suggested that momentum is a result of investor overconfidence and self-attribution. This research differs in its approach by using residual media coverage as a proxy related to the first signal in the model setting and by designing various tests to examine how investor overconfidence affects the magnitude of the media-based momentum.
The additional tests reveal three subsets of stocks that exhibit strong media-provoked momentum and subsequent reversal. The first is the subset of stocks for which the initial media signal is compounded by the tone, such that a winner stock may be further propelled by media with a positive tone (and vice versa). The second subset consists of difficult-to-value stocks that are considered to exhibit higher levels of uncertainty. The third subset consists of those stocks headquartered in highly individualistic US states.
How Did the Authors Conduct This Research?
The initial sample is made up of all common stocks traded on the NYSE, AMEX, or NASDAQ appearing on both CRSP and Compustat between 1989 and 2010, with adjustments for small illiquid stocks and bid–ask bounces. LexisNexis is used to obtain newspaper articles for all companies from the New York Times, USA Today, the Wall Street Journal, and the Washington Post as well as from local newspapers; only articles with a LexisNexis relevance score from 80% to 90% are used in the baseline tests. The availability of local newspapers was limited via LexisNexis during the beginning of the sample period. Approximately 2.2 million newspaper articles are assigned to more than 7,800 different firms in the dataset. Media coverage is defined as ln(1 + Coverage of articles) in a given month. The average tone of the articles is quantified by textual analysis, following the dictionary approach developed by Loughran and McDonald (Journal of Finance 2011), which considers the fraction of negative words in an article.
Multivariate regressions are run to explain media coverage; the explanatory firm variables include firm size, S&P 500 Index or NASDAQ membership, analyst coverage, and other specifications to control for various effects in order to obtain an adjusted R2. Momentum portfolios are constructed assuming a formation and a holding period of six months, respectively, with a month in between. Five equal-sized portfolios are created based on the residual national media coverage during the momentum formation period. In each quintile, the winner portfolio represents those stocks with a return above the 70th percentile and the loser portfolio represents those stocks with a return below the 30th percentile during their formation period.
Abstractor’s Viewpoint
Future momentum patterns may not follow the findings of this study because it is limited to newspaper coverage, but the research is useful in quantifying the impact of media on momentum during a time when newspapers were used more for investment news compared with today. I would be interested to see further research to take into account the impact of newer media sources, such as Twitter, on momentum.