From the January 01, 2010 issue of Futures Magazine • Subscribe!

Trading portfolios: Diversification and correlation spikes

“Paul,” (not his real name), a former floor trader, has a problem. This didn’t used to be a problem when he simply traded corn, but now he trades markets in several different sectors. It is safer that way, or so he thought.

“When I was on the floor, I traded one thing at a time. I didn’t have to worry about who was correlated to what,” states our trader. “If I’m trading corn and every damn market in Christendom drops like a rock what do I care? I care if I’m long corn, but otherwise no. Now, I’m upstairs and if everything starts going in the same direction I’m in big fat trouble.”

He also can see some big profits, which is the ideal time to adjust his position sizing to move in step with the new reality. Many managers were able to do that in the second part of 2008 when they realized that their outsized profits in nearly every sector they traded could be turned against them in a hurry. Several of the Top Traders of 2008 (see March 2009), many of whom earned in excess of 100%, reported reducing position sizes due to this correlation. This helped reduce the effect of the inevitable correction that came in the first quarter of 2009.

The big fat trouble “Paul” could be in is being on the wrong side of almost everything. Most of the time, most markets move more or less independently. There are, of course, higher correlations within sectors and soybean prices may respond to changes in the price for gasoline, but they respond to a lot of other things, too. Anyone trading both soybeans and gasoline futures, for example, is making two more or less independent bets and that’s a good thing. But every now and then much of the economy will be dominated by just one thing: the collapse of the banking sector, a plane crashing into the World Trade Center, a natural disaster. When something like that happens, price correlations spike and most prices move up or down together (see “So you are diversified,” right). When that happens a fund manager is no longer trading a portfolio. He is no longer making lots of independent bets. He is making one bet and if he loses, he loses big.

These days, most commodity trading advisors (CTA) and fund managers trade a lot of different markets simultaneously. So do most sophisticated traders. Many funds and the majority of CTAs trade in as many different markets as they can and use the powerful tools of portfolio theory to manage their risks. All of this works reasonably well most of the time and sometimes it works wonderfully — as long as most of the trades are not highly correlated and as long as the correlations remain fairly stable.

There are a lot of different measures of correlation and all of them measure things a little bit differently. The measure most fund managers use is called the Pearson correlation coefficient and it measures correlation on a scale from one to minus one inclusive (See “The Pearson method,” page 54). A correlation of one, also called a perfect correlation, means that every movement of market A is accompanied by a similar movement of market B. If there is a 5% rise in market A, it will be accompanied by a 5% rise in market B. Strictly speaking, every movement in market A doesn’t have to be duplicated by market B. The correlation can be perfect if every movement in market A is, say, exactly doubled by market B. The correlation is also perfect if, say, every movement in market A is matched by three and a half times the movement in market B plus an extra percent.

In a perfect negative correlation, every movement in market A is matched by an equal opposite movement in market B. Every rise in market A is accompanied by a drop in market B and vice versa.

No correlation means that markets A and B move independently; they have nothing in common.

Market correlations are almost never negative and never perfect and almost never close to being perfect. In the real world, as long as correlations remain fairly stable; that is, a correlation of, say, 0.4 in 2008 remains fairly close to 0.4 in 2009, diversification works reasonably well. But while correlations are often reasonably stable, they vary over time. And sometimes correlations wander far from their historical norms. So, two markets whose correlation coefficient was 0.5 in 2008 might suddenly have a correlation coefficient of 0.15 or 0.93 in 2009. Drops don’t matter. But when everything or almost everything is headed in the same direction the portfolio is suddenly much riskier than its owner designed it to be. In which case, the fund manager might well make or lose much more money than he expected.

So, what can a trader do?

comments powered by Disqus