Do non-professional analysts who publish on Seeking Alpha offer valuable stock-picking advice? In their August 2020 paper entitled “The Cross-Section of Non-Professional Analyst Skill”, Michael Farrell, Russell Jame and Tian Qiu measure skill among such analysts as the hypothetical abnormal return an investor would earn by following reports/recommendations that focus on one common stock over 5-day or 63-day post-publication holding intervals. They classify recommendations as buy or sell using either: (1) disclosed author positions, or (2) sentiment of associated reports inferred from word usage. They measure abnormal return for each recommendation as its 6-factor alpha, adjusting for market, size, book-to-market, profitability, investment and momentum factors calculated from daily returns from 13 months to one month before the recommendation. They further test an implementable trading strategy that buys (sells) at the ask (bid) and subsequently sells (buys) at the bid (ask) price at the end of the holding period, with and without delays of 24 to 72 hours after publication. Using 123,120 Seeking Alpha research reports prepared by 1,879 non-professional analysts (each with at least 10 qualifying reports) and focused on single common stocks, along with contemporaneous stock and factor returns, during 2005 through 2017, they find that:
- About 45% (30%) of sample reports are positive (negative).
- Average 6-factor alpha across all analysts in the sample is 0.38% (0.68%) over a 5-day (63-day) holding interval, with positive alphas for about 60% (71%) of analysts.
- There is considerable dispersion in skill across analysts amounting to about 10% annualized alpha.
- A strategy that follows equal-weighted recommendations of analysts in the top fifth (quintile) of alphas over their respective last 10 reports generates 1.78% gross monthly alpha, more than double the 0.79% gross monthly alpha for all analysts. The significant outperformance of the top quintile is robust to different risk adjustments, different measures of past performance, different holding periods and exclusion of reports that coincide with major firm information events such as earnings announcements, release of professional research reports and media coverage.
- After using bid-ask spreads as trading frictions and imposing a 72-hour lag after report publication, following recommendations of top-quintile analysts generates 0.84% monthly net alpha (about 10% annualized), compared to -0.09% gross monthly alpha for all analysts.
- Analysts who are topically specialized and/or who actively comment on other research reports tend to outperform other analysts.
- Publication of recommendations by top-quintile analysts does not induce unusual order imbalances in associated stocks, helping to explain why following the recommendations with a delay works.
In summary, evidence indicates that the non-professional analysts who make the best individual stock recommendations on Seeking Alpha exhibit potentially exploitable skill.
Cautions regarding findings include:
- As noted in the paper, trading frictions do not account for costs other than bid-ask spread, such as transaction costs and shorting costs/constraints.
- Returns represent trade-level rather than portfolio-level performance. Because the number of active recommendations varies, an investor would have to keep idle capital available for new opportunities over time, thereby necessitating a cash reserve that would lower overall return. Maintaining equal position weights would be challenging for such a portfolio.
- The test approach is beyond the reach of most investors because it involves:
- Continuously monitoring thousands of reports and hundreds of stocks to identify top analysts.
- Holding a portfolio that is, on average, long 129 stocks and short 97 stocks as recommended by top-quintile analyst recommendations.
