Beware the Bangladeshi Butter Bias

Bangladeshi butter.

If you’re looking for the indicator with the highest correlation to the S&P 500, it’s not GDP growth or even earnings growth. No, it’s butter production in Bangladesh.

In a paper published two decades ago, mathematician David Leinweber and portfolio manager Dave Krider found that butter production in Bangladesh had the tightest correlation to the S&P 500 of any data series they could find.

Collectively with American cheese production and the Bangladeshi sheep population, this three-variable model “explained” 99% of the S&P 500’s movements.

Now, before you roll your eyes, Leinweber and Krider published the study in jest. These are geeky quants enjoying geeky quant humor. It’s precisely the kind of thing I would have done with my grad school buddies after a few pints at the Three Tuns in London. We would have gotten a good laugh out of this wild idea and then likely gotten a drink dumped on our heads after trying to ask for a girl’s phone number.

The problem is that a lot of traders aren’t smart enough to know it’s a joke.

In an article he wrote for Forbes a few years ago, Leinweber commented that 20 years later he still gets calls to his office asking for the latest Bangladeshi butter production figures.
Amazingly, it’s not just knuckle-dragging traders that fail to get the joke. Even professional researchers fall for it — some more than others.

Leinweber recounts a 2011 paper written by Tatu Westling, a professor at the University of Helsinki, that “explores the link between economic growth and penile length.”
Yes, you read that right.

According to Westling:

The average [GDP] growth rates from 1960 to 1985 are found to be negatively correlated with penile lengths: a unit centimetre increase in its physical dimension is found to reduce GDP growth by 5% to 7% between 1960 and 1985. Quite remarkable is the finding that male organ alone can explain 20% of the between-country variation in GDP growth rates between 1960 and 1985.

Regarding the relative importance of political institutions in shaping economic development, it seems that male organ is more strongly associated with GDP growth than country’s political regime type.

So, for every additional centimeter of manhood, GDP growth falls 5% to 7%.

As with all things in life, it appears there are trade-offs. If I have to choose between the two, perhaps wealth is overrated.

The author goes on to hypothesize that well-endowed men are inherently happier and more confident and thus don’t feel the need to work as hard, thus the lower level of economic development.
Apparently, the massive skyscrapers that penetrate the New York skyline really are a form of compensation.

Professor Westling very well might have written the paper in jest. I sincerely hope so.

But you don’t have to look far to see plenty of nearly equally ridiculous market indicators. The Super Bowl Indicator, the Hemline Indicator, head-and-shoulders patterns… I could go on all day. I’ll spare you a long explanation on overfitting, spurious correlations and data mining.

For crying out loud, I’ve spent the last half of a page making penis jokes, so getting overly technical at this point would be ridiculous.

But I can summarize it like this: If you can’t plausibly explain why a model works, then chances are good it’s a random coincidence.

There is absolutely no plausible reason why Bangladeshi butter production would predict S&P 500 returns and it’s hard to argue that average penile length determines GDP growth. (See what I did there?)

So, when looking at a model, ask yourself: Does it intuitively make sense?

As an example, I spent most of last year backtesting a stock trading model that combined several value investing metrics with momentum metrics.

The result was a portfolio strategy that delivered 43% backtested annual returns from 1999 to the present, utterly crushing the S&P 500.

Now, could this be an example of overfitting the data?

Maybe. But I don’t think so because the approach intuitively makes sense. Countless studies (most famously Fama and French’s 1993 study; click here to download it) have shown that value investing outperforms over time, and most of the famous investors throughout history, such as Warren Buffett and Benjamin Graham, were value investors.

Value investing works.

Meanwhile, various momentum and trend-following strategies have also been proven to outperform over time. So, it stands to reason that combining value and momentum would lead to solid results. I can’t guarantee that the future live results will live up to the backtest. You know the drill: Past performance is no guarantee of future results. But I’m betting it generates better returns than a Bangladeshi-butter-based market timing model.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.