Josef Lakonishok and Robert Haugen wrote a book, The Incredible January Effect, in 1988. At about 110 pages, it's a quick read, and you can get a used copy on Amazon for under $10 delivered to your door (I love used books). They presented the then most prominent behavioral anomalies at that time. This highlights the problem of the behavioralists.
The book starts with an anecdote about how some bankers noticed a pattern in interest rates, and the Efficient Markets/Random Walk blinders made this impossible for these professors to see. Interest rates are low in March, high in October, a difference of about 50 basis points! Alas, this effect, I must say, I was unaware of. It certainly does not show up in the Fed's H15 data. As neither author talks about this anymore, I think this is just one of those little mistakes everyone hopes people will ignore. So I will.
Their big idea is that basically all of the 'risk premium' in stocks and bonds occurs in January. This is because institutions pick stocks then, pushing up prices of risky, and small stocks (which are risky). Over the year stock pickers stop 'buying', so stocks just meander the other 11 months of the year. Alas, most of this effect turned out to be driven by low-priced, illiquid stocks. These small stocks would move from 3/8 to 1/2, back to 3/8, and the daily cumulative return was huge, because the tapes showed +33%, -25%, +33, -25%, etc. Add to this that your retail prole got a horrible fill back in those days (this is when they invented the term 'rip your face off'), so you would actually be adversely filled, so the only one receiving the January effect where the floor traders. playing this game just generated excess trades. No one thinks the January Effect is real anymore, it was an artifact of crappy, high frequency data on a biased sample (illiquid stocks).
This same bias drove the famed Thaler and DeBondt mean-reversion trade, which conspicuously has no children in the literature (see Conrad and Kaul, 1993). To the extent their is negative autocorrelation at 36 months, it is imperceptible. In contrast, the momentum effect of Jegadeesh and Titman, which finds positive autocorrelation at 12 month horizons, is robust to excluding low-priced stocks, and has remained in the literature.
Currently people are focused on biases from excessive complexity of asset backed securities, and the spreads for all such assets has risen as they have all been tarnished by the poor performance of AAA rated mortgage-backed securities. Yet this 'bias', seems rather focused on one bad assumption, that housing prices would not fall. It really was not much more complicated than that. Regulators and the market are punishing these products. I don't see this as a big improvement, in that excessive pessimism, like excessive optimism, is wrong, and the similarity is merely the structure. This is a flawed diagnosis.
In general, behavioral explanations cover so many scenarios—optimism, pessimism—one has to have a specific behavioral insight to make an reasonable argument. Otherwise, you are like someone in favor of a 'third party' as an alternative to our two-party system, a group with less in common among itself than with the parties it is an alternative to.
I haven't spent $52 on a book since grad school (actually I don't think I've bought more than 10 books since grad school, and at least half were about babies) but I at least requested that my public library buy it.
"Their big idea is that basically all of the 'risk premium' in stocks and bonds occurs in January. This is because institutions pick stocks then, pushing up prices of risky, and small stocks (which are risky). Over the year stock pickers stop 'buying', so stocks just meander the other 11 months of the year"
I don't see how this is a behavioral story. Either institutuions bought all their stock January back then (empirical question) or they did not. If so seems like reasonable rational story.
"These small stocks would move from 3/8 to 1/2, back to 3/8, and the daily cumulative return was huge, because the tapes showed +33%, -25%, +33, -25%, etc."
If you value-weight in calculating returns (which they may have done), this won't be an issue.
"To the extent their is negative autocorrelation at 36 months, it is imperceptible"
Long term reversal is well documented. You can see returns here:
for large stocks that avoid errors from low-volume stocks, the mean-reversion had an annual return of 3%, and a vol of 19%, which I think is 'small'.
Haugen and Lakonishok, anyway.
Post a Comment