Mathias Hasler writes:
I have a working paper about the value premium and about seemingly innocuous decisions that are made in the research process. I wanted to share this working paper with you because I think that you may find it interesting and because your statistics blog kept me encouraged to work on it.
In the paper, I study whether seemingly innocuous decisions in the construction of the original value premium estimate (Fama and French, 1993) affect our inference on the underlying value premium. The results suggest that a large part of the original value premium estimate is the result of chance in these seemingly innocuous research decisions.
Here’s the abstract of the paper:
The construction of the original HML portfolio (Fama and French, 1993) includes six seemingly innocuous decisions that could easily have been replaced with alternatives that are just as reasonable. I propose such alternatives and construct HML portfolios. In sample, the average estimate of the value premium is dramatically smaller than the original estimate of the value premium. The difference is 0.09% per month and statistically significant. Out of sample, however, this difference is statistically indistinguishable from zero. The results suggest that the original value premium estimate is upward biased because of a chance result in the original research decisions.
I’m sympathetic to this claim for the usual reasons, but I know nothing about this topic of the value premium, nor have I tried to evaluate this particular paper, so you can make of it what you will. I’d be happier if it had a scatterplot somewhere.
I don’t do finance, but I remember hearing from someone that Fama himself told people that his model was basically an exercise in data mining.
That paper made the rounds last year. Certainly interesting, but that there may be an issue with forking paths does not necessarily invalidate the entire factor modelling approach. Some folks in connection with AQR have a paper [1] on how most of the factor modelling research replicates.
[1] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3774514
True, but other people not associated with AQR (or, as far as I know, with any other institution that would expose them to conflicts of interest) have a paper on how most of the anomalies don’t replicate: http://global-q.org/uploads/1/2/2/6/122679606/houxuezhang2019rfs.pdf
Thanks. I recall that one too. I would need to re-read them both to make a better determination on which is better.
There is a gigantic literature on the size of the risk premium and I think it’s fair to say that there are reasonable estimates all over the place. You’re trying to measure a latent thing whose realized value jumps around all the time, so it’s unsurprising that nailing down “the” risk premium is impossible. It’s certainly a matter of some historical interest that the *first* and *most famous* estimate of the risk premium is too high. But we know that the initial estimates of almost everything (e.g. early childhood interventions) are too high, for reasons you’ve discussed many times. That that “too highness” can be traced to a few reasonable forking path decisions is actually a good sign for the literature in general. But there’s not much else that could have caused it, because there is no shortage of data, and the result has been replicated a gazillion times.
Jonathan:
Not only were the initial estimates of the effects of early childhood interventions too high; the more up-to-date estimates continue to be too high!
Coincidence methinks,
https://www.linkedin.com/posts/peter-clark-b6663b5a_ifrs-standards-use-too-many-different-terms-activity-6949758001658441728-stOt
Then, gotta digress – I had CBDC in mind reading around about these things…