Do human or algorithms perform better at picking stocks?
getty
FinTech has seen an explosion in tech-driven financial service offering with over $50 billion invested as of 2018. But does it work? Do seasoned investment analysts or algorithms make superior stock picks? Researchers at Indiana University have recently examined this question.
The research was conducted by Braiden Coleman, Kenneth Merkley and Joseph Pacelli examining 76,568 robo-analyst reports over the 2003-2018 period. This is what they found.
Key Differences
More Balanced Recommendations
One criticism of human analysts is that they seldom issue sell recommendations on companies. About 1 in 20 ratings by a human analyst is a sell. For robo-analysts that figure is 1 in 4. This is perhaps because human analysts want to maintain a relationship with the company they cover, and to potentially nurture investment banking relationships. Whereas robo-analysts are parsing through SEC filings and similar datasets rather than meeting with management. Of course, this doesn’t make robo-analysts better, they just distribute rankings more evenly as you would expect from a algorithmic process. However, it does suggest that robo-analysts are less prone to excessive optimism, which can be an issue for human analysts.
More Revisions
Robo-analysts tend to make more revisions to their recommendations. As the news changes they alter their opinion. Again this alludes to a behavioral bias related to anchoring that may impact human analysts. Once you have stated a position publically, it can be hard to change it. Again, robo-analysts don’t fall prey to this potential issue. That said, more frequent changing of recommendations can drive trading costs and tax issues for a firm following those recommendations. So more frequent updates may not always be a positive thing for a portfolio.
Different Windows
Robo-analysts tend to update their analysis at different times. Given robo-analysts are churning through SEC data, updates tend to follow material data disclosure such as 10Qs and 10ks. In contrast, human analysts are more likely to update following earnings announcements when management presents slides and answer investor questions. Both of these are usual sources of information, but clearly human analysts focus more on one, and robo-analysts on the other.
Investment Wisdom
The above differences are interesting, but the key question is who makes more money with their recommendation? The researchers found that robo-analysts buys beat humans in their returns by around 5% a year over the 15-year research window. That’s quite an edge. Especially since human analysts do tend to move the market when they issue a recommendation, whereas lower profile and perhaps less highly regarded robo-analysts tend not to.
However, there’s a catch. Robo-analysts were worse in their sell recommendation than humans. As a result, the picture is nuanced. If the robo-analysts were truly superior, then both their buys and sells would beat humans. That’s not the case. It appears that robo-analysts are definitely useful, but not always superior.
What The Future Holds
It appears that both human and robo-analysts, in aggregate, can yield investment insight. Indeed, the methodologies appear complementary and, of course, so-called robo-analysts do rely on human judgement to create, maintain and review the algorithms used. However, as FinTech continues to rapidly evolve as techniques and data sources improve, robo-analysis will have an increasing role to play in the coming years. That they have outperformed human analysts over recent history is impressive considering the future innovations that may be in store.
