“Your masters let you play with dangerous toys.”
Ancient Greek toy horse. Photo: Sharon Mollerus. Source: Wikipedia
That’s a line from a work of science fiction written over seventy years ago – before star-fighters and warp drives. Two characters confront one another for the first time, and one believes that the other doesn’t know what he’s talking about. They do end up cooperating, though, and the forces they harness are incredibly powerful.
That’s the way I often feel about academic finance. Researchers blithely discuss factor investing and debt overhang and capital structure, but I get the sense that most of these folks haven’t done much more to invest other than contribute to their Universities’ 403(b) plans, much less ever advised someone else. There’s a sort of breezy indifference to the implications of their work – as if statistical significance matters more than client outcomes.
And the results can be quite influential – like option pricing or corporate governance. Who serves on a company’s Board of Directors, for example, can have a huge impact on the lives of thousands of people. But for every study that points one way, there seem to be others headed in the opposite direction – like medical researchers who find that wine, coffee, and beef both cause and prevent cancer.
Source: AJCM, Vox.
One analyst estimates that only 6% of peer-reviewed medical research is well-designed or relevant enough to inform patient care. And this is understandable: researchers need to “publish or perish,” and editors are busy with their own projects and priorities. It’s no surprise, then, that a prominent editor of a prestigious finance journal accused his profession of data-mining: dredging through studies and results until a statistical relationship can be found, then concocting a narrative to justify the correlation. He calls this practice “p-hacking”: evaluating the data from different perspectives until a suitably high t-statistic – and low p-value – can be produced (and published).
Example of spurious correlation. Source: Wikipedia
Researchers are people too, and journal editors compete for citations and web-traffic. Public choice theory shows government officials respond to incentives, and this also applies to academics. Perhaps this is why debates in academic finance seem like those of scholastic monks determining how many angels can dance on the head of a pin. They need to get their findings into a top journal in time to impress the tenure committee.
So I don’t get too worked up when the “dangerous toys” of statistical analysis seem to predict that firms with meaningful stock ticker symbols or companies with Board members who own cats tend to do better. Such studies won’t change our practices. It just seems like a lot of research should come with a warning label.
Douglas R. Tengdin, CFA