Feb

14

 There was a lot of interest a few years ago in Joel Greenblatt's book The Little Book That Beats the Market . He set up a web site www.magicformulainvesting.com which selects stocks based on this method (using a combination of 2 criteria: earnings yield and return on capital).

Two years ago I saved a list of 30 stocks it recommended and just recently I calculated the performance up to now. The results are below:

MagicFormula Investing Stock total return from 2012/01/29 to 2014/01/29

Source: Bloomberg

          %Ret
atvi       43.04
apol      -40.51
amat       44.77
hrb        87.55
cub         9.80
dell      -14.40   acquired 10/30/2013
dlx       102.63
dlb        27.57
xls       110.55
expe      111.76
gtat       10.06
gme        54.34
esi       -44.16
icon       95.89
idcc      -18.98
klac       24.79
lps       129.00   acquired 1/3/2014
lo         51.17
mant      -10.67
mrx        37.67   acquired 12/11/2012
msft       32.90
nsu       -38.53
noc       103.22
prx        37.99   acquired 10/01/2012
rtn        94.84
ldos       48.21   ticker change from sai to ldos
sndk       48.91
save      183.80
stra      -65.67
vphm       67.37

average    44.16

spy        40.30
rsp        44.02

As you can see the return over 2 years was good, at 44.16%, but indistinguishable from the equal weighted S&P at 44.02% and close to the (ordinary) S&P at 40.30%.

[Strictly speaking you are supposed to hold the portfolio for a year and then rebalance it, which I forgot to do, so it is not a completely accurate replication of the MagicFormula method.].

Bill Rafter writes:

Before Greenblatt there was Haugen who ranked stocks by 60+ variables. He published them in his book, the Inefficient Stock Market. BTW all of his books are interesting. Haugen's material looked really good, but seemed to fail to deliver outstanding returns.

Then Greenblatt's book came out and showed good results from only using two variables: ROE and ROA. One of those was a Haugen variable (I forget which). Then Haugen produced a second edition in which he bumped up the number of variables to ~70, and included both ROE and ROA.

We did a lot^2 of work with both H and G and found that neither are particularly good at picking winners, but both are very good at finding clunkers. And of course if you eliminate the clunkers, the remaining portfolio will certainly outperform the market indices. But of course that only really works if you essentially invest in ALL of the non-clunkers. Thus they are good strategies for large mutual funds.

FWIW we found that you can cull out most of Haugen's variables (but include ROA and ROE) and do fairly well with 16 variables. But again, they are best used as ranking tools to eliminate the underperformers. Our opinion is that both H and G methods are the financial markets equivalent to comfort food. If comfort food makes you feel better, then use it.

anonymous writes:

I agree that Haugen (1942-2013) was an interesting character who did some innovative work. I have read many of his books/papers; how well it holds up remains to be seen. He used the kitchen sink approach which Rocky so detests, but he found things (such as the low vol phenomenon) that were new and interesting (now low vol has maybe become too popular). BTW the two year test period I used is far too short, it could be argued. You need more like 10 years.

Gordon Haave writes:

"Our opinion is that both H and G methods are the financial markets equivalent to comfort food."

This by the way is very important for non-professionals. Most not-professionals lose money because of over-trading, moving in and out on a whim, chasing winners, etc. (actually, most professionals do as well).

A system like Greenblatt's that tells a good story, tells you to only re-balance one per year, and does "as good" as the market will tend to outperform what the individual otherwise would have done.

Richard Owen writes:

Bill, when you say clunkers, do you mean it was great at selecting clunkers by the lowest ranking? Or it was great at selecting clunkers by false positives - i.e, many of the best stocks were very bad despite their attractive ROE etc. and this is what hurt performance? And so they are good for 'cleanup'? Perhaps it's a bit unfair to pick '12-14 data, but very interesting nonetheless. I suspect the best outperformance of such strategies would come in the early bull market and then converge with the market as it matures, then lagging at the end / collapse of the bull?

Bill Rafter writes:

Richard,

Sorry, I did not mean to be cryptic. By clunkers I meant stocks that you should not own, because they are and will be poor performers.

We used both Greenblatt and Haugen variables to rank the Russell 3000 for 20 years and put the stocks into 10 bins (i.e. deciles) and observed the performance of those bins over the following 6 months (with no overlapping periods). Considering the deceased stocks the total number of securities exceeded 8,000. You have to include the dead stocks or you will have survivor bias. The lowest ranked bin performed really bad, such that they could have been shorted profitably (we are long-only and had no interest). Importantly the performance of each higher decile was monotone increasing. That last bit means the strategy has merit. However, the top three deciles did not have the performance that would encourage us to pursue the strategy. That is, they outperformed the market, but they were not an improvement over other strategies.

It makes perfect sense to find those clunkers and eliminate them from your subsequent ranking routines. But there are many ways to find them. It turns out that the underperforming stocks have the mark of the beast written all over them, and you don't need to look at 10+ variables. If data mining for fundamentals is too onerous for the researcher, ranking by volatility will find clunkers. Deselecting for volatility will also keep you out of some high-flyers, but that's a good price to pay. High short interest coupled with declining prices is another.

We did the research 2-3 years ago and I apologize that I don't have the time to dig it out and clean it up for public consumption, as it was only done for our shop. Realize that we are statistical traders who typically hold either ~20 or ~50 positions because we cannot forecast individual security performance, but only bin performance.


Comments

WordPress database error: [Table './dailyspeculations_com_@002d_dailywordpress/wp_comments' is marked as crashed and last (automatic?) repair failed]
SELECT * FROM wp_comments WHERE comment_post_ID = '9107' AND comment_approved = '1' ORDER BY comment_date

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search