The author examines whether the aggregated opinions of consumer crowds, consolidated from online product reviews at Amazon.com, serve as a new form of fundamental information on the supplying firms, conferring predictive power about future stock returns. His findings could be of interest to investors in signaling a timely quantitative method for screening attractive market opportunities.

For example, a spread portfolio constructed to purchase stocks with abnormal favorable customer ratings in the top tercile and selling stocks in the bottom tercile generates an abnormal monthly return of 55 bps to 73 bps. Furthermore, these ratings might be useful in positively predicting revenue and earnings surprises, as well as net purchases by hedge fund managers of the involved stocks.

These results are concentrated among stocks in less efficient market sectors—for example, those that are small cap and/or with low analyst coverage. The results appear to hold after controlling for such potentially determinative firm characteristics as profitability or prevalence of advertising. The estimated return predictability holds over the long run, and it can be available sooner than projections via such traditional information channels as equity analysts. In examining the “wisdom of crowds” content of consumer opinions, the author distinguishes between an information story (fundamental information conveyed) and an attention story (naive reaction to observed behavior of others) and shows that the opinions offer additional substantive guidance about firms beyond that reflected in stock prices.

During July 2004–December 2015, 14.5 million customer reviews of 270,000 different products were posted at Amazon.com. The author assesses 346 public firms, with the top industries being business equipment, consumer nondurables, and manufacturing. This sample is much larger and more comprehensive than those of previous researchers of online consumer feedback.

The panel dataset covers firm months, with each month having at least 10 reviews and, on average, including 150 stocks. An average monthly rating (on a scale from 1 to 5) for a firm’s products is computed and compared with the average rating for the preceding 12 months, the difference being the derived abnormal rating. Baseline financial data on stock returns and volumes, accounting ratios, and analyst earnings estimates are drawn from CRSP, Compustat, and I/B/E/S. Measures are derived for revenue and earnings surprises and related hedge fund trading.

Regressions assess the determinants of the abnormal ratings—examining the ratings’ investment value, their predictive effect for future returns over the long run, and their status as new information applying Fama–Macbeth tests—and estimate the ratings’ relationship to cash flow surprises and trading by sophisticated investors.

Investors may acclaim the identification of a new and potentially effective interpretive model for decision making. The author demonstrates that the empirical influence of a significant contemporary industry disruptor can extend beyond its immediate marketplace, which could very well resonate as part of the internet zeitgeist.

Reality unfortunately always offers caveats. Anecdotal evidence is accumulating that customer postings are curated online, particularly by smaller companies striving to put their best foot forward via fictitious reviews. Amazon has policies against such shenanigans, which represent unfair trade practice. But gains from cheating can be tangible, and prevention of these practices is likely modest at best. Time will tell as to how long these highlighted predictability effects, or arbitrage profit opportunities, can endure.