Can Analysts Pick Stocks for the Long Run?

Academic research has uncovered a persistent drift in stock prices following (and in the same direction as) forecasted earnings revisions from financial analysts. The post-revision drift (PRD) anomaly implies that investors are slow to process earnings news in analysts’ forecasts. It is also one of the more puzzling anomalies in finance. One hypothesis for the PRD anomaly’s existence is that transaction costs can be high enough to prevent arbitrageurs from correcting mispricings, allowing it to persist.

Oya Altinkilic, Robert Hansen and Liyu Ye contribute to the literature on the PRD anomaly with their study, Can Analysts Pick Stocks for the Long-Run?, which was published in the February 2016 issue of the Journal of Financial Economics. Their study covered the period from 1993 through 2010. Controlling for the well-documented post-earnings announcement drift (PEAD), which could enlarge measurements of average PRD, and splitting the full period into a pre-period (1993 through April 2003) and post-period (May 2003 through 2010), the authors found:

• On average, PRD is no longer persistently different from zero.

• Investors are no longer underreacting to revisions of earnings forecasts.

• There’s no longer support for the hypothesis that analysts typically supply new information that correctly picks stocks for the long run.

• There also was no significant association with extreme revisions, a commonly used proxy for better-informed analysts.

• PRD has broadly vanished because of a general decline in transaction costs, pushed down to historic lows by decimalization, the expanded use of supercomputers and algorithmic trading. The disappearance of PRD coincides with notable reductions in transaction costs that have attracted profit-taking arbitrageurs to PRD.

• Consistent with the hypothesis that high transaction costs allow anomalies to persist, there is evidence of statistically significant average PRD in the lowest decile (the bottom 10% of the revisions) when ranking by trading volume. Stocks exhibiting PRD tended to be very low priced, have limited trading volume and are more likely to be listed on the Nasdaq, thus having higher trading costs (median quoted bid-ask spreads are larger in the supercomputer era for Nasdaq-listed firms than for NYSE-listed firms).

• Out-of-sample tests covering the period from 1996 through 2010 confirm a general absence of PRD during the post-period in the other six countries studied: Canada, France, Germany, Italy, Japan and the U.K., countries which have also had declining trading costs.

The authors’ findings are consistent with the themes Andrew Berkin and I discussed in our book, “The Incredible Shrinking Alpha.” In it, we presented evidence of a persistent trend that has resulted in an ever-declining percentage of active managers able to generate statistically significant alpha. In 1998, when Charles Ellis published his book, “Winning the Loser’s Game,” about 20% of actively managed mutual funds were generating statistically significant alpha. Today, that figure is about 2%, even before considering taxes.

As the authors noted, the implication for investors is that “the general decline in transaction costs has allowed information to be incorporated more quickly and completely into security prices, eliminating both profit opportunities from strategies that use analysts’ revisions, as well as the predictability of long-run returns based on the revisions. The disappearance of PRD therefore exemplifies how rational pricing of securities interacts with changes in real inefficiencies to extend market efficiency.”