Statistical Value 2014

When I think “value investing” I think “Bruce Greenwald”. And when I think “Bruce Greenwald” I think of an investing approach whereby one deeply studies individual companies in order to decide what to invest in: analyzing competitive advantages, playing “what if” scenarios with competitors via game theory, and of course performing valuations.

But as detailed in Greenwald’s Value Investing book, there is clearly another way to skin the cat, and that is to employ mechanical strategies of buying baskets of companies that statistically put you on the right side of the trade. This he refers to as statistical value.

In the book, Greenwald provides 3 approaches one may use to go about this. I call these Operational, Multiple, and Technical.

OPERATIONAL involves screening companies with operating metrics one might use to identify a good business. We’re talking ROE, profit margins, sales growth. Except… you’re not screening for excellence. You’re screening for disaster. The idea is to pick the very worst to profit from mean-reversion as, to quote Horace, “Many shall be restored that are now fallen, and many shall fall that are now in honor.”

MULTIPLE involves buying the companies with the lowest valuation ratios. Lowest P/B, lowest P/E, lowest P/CF ,etc. – whatever multiple floats your boat since historically they’ve all evidently worked. Again, you don’t want the ones that are slightly cheap. To profit from others’ irrational selling you have to buy the icky lowest decile (and short the highest too, if it suits you).

TECHNICAL involves buying the companies with the worst 3-yr trailing price returns. These are the stocks investors are dumping not due to a bad quarter or two, but because they’ve been breaking their hearts for a few years now, and they just can’t stomach holding them any longer.

As mentioned in another post, the overwhelming temptation to those who decide to do statistical value is to use one of these screens as a starting place, but then cherry-pick the companies they like best, or throw out the ones they’re sure won’t work out. And the overwhelming evidence apparently is that when you do this, loss aversion bias kills you right out of the gate!

And what would you have gotten historically with plain ol’ unmodified statistical value? About 3-4% higher return than the market over the long term, that evidently puts you in the top 2% of money managers.

So this year, I’m beginning to test-drive it. I took the 1500 companies in the S&P 400, 500, and 600 and sorted by 3-year total return. I was going to just invest in the worst 10, but I noticed that if I picked my 10 from the worst 14 stocks I could simultaneously get all US companies and lowest decile P/B too.

At the opening bell this morning, I sold the 5% of my portfolio that was in Wal-mart and put 0.5% each into the following. I know! 5%?! Very brave, right?! Let me just tell you – knowing the problems that some of these companies face, and seeing the $ value I was taking out of stable WMT to do this, did NOT make it easy.


In academic studies, when you buy the worst performers of the last 3 years, you usually let 3 years pass before you tally how you did. So check back on January 2, 2017 to find out whether I’m a contrarian genius or a total idiot.

4 thoughts on “Statistical Value 2014”

  1. Year to date as of 8/27/14:

    S&P 500: +9.4%
    The 10 Statistical Value positions: -6.9%
    WMT (which I sold to fund them): -1.8%

    It’s not supposed to be until years 2 and 3 that the outperformance starts to materialize. Greenwald says 2/3 of the positions will do terribly but the 1/3 that do well will more than compensate. So far, only 2 of the 10 have been outperformers (JCP and NFX).

Leave a Reply

Your email address will not be published.