Felix Salmon

Many thanks to Scotty Barber for putting this chart together for me. It shows the extraordinarily high correlations that we saw within the S&P 100 at the height of the Lehman crisis; at the height of the Greece crisis; and then, again, for pretty much the entire second half of 2011. At that point, high correlations really did look as though they were the new normal.

Obviously, correlations within and across different asset classes don’t always move in tandem with each other. But in general, the RORO trade, as it’s known, (risk-on, risk-off) tends to send correlations soaring across the board. And I can’t help but wonder whether that huge plunge in correlations that we see at the beginning of 2012 was related to the way in which JP Morgan’s CIO blew up.

Remember that the CIO’s main job was to make hedges: buy buying or selling one thing, the idea was that the bank would protect itself against losses on some other thing. So in order for hedges to work, those two things need to continue to be highly correlated.

But if you look at this chart, the period when Bruno Iksikl was putting on his huge CDS index trade was also the period when correlations were falling at an almost unprecedented pace.

Jamie Dimon, from the day he revealed the losses, has had nothing but the harshest possible words for the hedges in question, saying that they were flawed and should never have been put on. But that’s easy to say in hindsight. Maybe they were really great hedges, in a high-correlation world — and then correlations fell apart, and the trades started moving against JP Morgan, and they had to get bigger and bigger in order to retain any hint of actual hedging capability. Obviously we don’t know for sure that’s what happened. But it’s certainly consistent with movements in correlations this year.

Paul Murphy has a good overview of RORO today: the risk-on, risk-off phenomenon whereby all assets are increasingly correlated. HSBC has even come up with a RORO Index, which, you won’t be surprised to hear, is going up and to the right:

HSBC’s correlation matrices are prettier. In 2005, the world of investable assets was lovely and turquoise, full of low-correlation asset classes. Today, not so much.

What really fascinates me about these correlation matrices, however, is the rankings. Look down the left-hand side, and you’ll see the big risk-on asset classes at the top, and the big risk-off asset classes at the bottom. Here’s how the lists changed between 2005 and today:

There are strong similarities along the general lines you’d expect: stocks are at the top, bonds are at the bottom. But the differences are fascinating — none more than the fact that the dollar has moved from being at the top of the list in 2005 to being at the bottom of the list in 2012. Right now, the dollar is the ultimate safe asset; it’s not always thought of that way.

And while there were was a cluster of currencies at the bottom of the list in 2005 — the Swiss franc, the yen, the pound — they’ve now moved up into the middle of the list, in a world where pretty much every currency in the world has zero interest rates. The carry trade, at least between developed-world hard currencies, ain’t what it used to be.

Meanwhile, the VIX, which was right at the bottom of the list in 2005, is there no longer. It’s still used to hedge against downside volatility. But over the past seven years it has become much more of a traded asset class in its own right, and so while it used to have really strong negative correlation with, say, the Nasdaq, it doesn’t any more.

All of which is to say that correlation itself is not a simple risk-on, risk-off thing. In general, correlations rise as the RORO index goes up. But specific correlations can change in unpredictable ways. Sterling and natural gas might be among the few relatively uncorrelated asset classes today. But there’s no reason to believe they’ll stay that way tomorrow.

Tadas Viskanta does a good job of rounding up the reaction to the news that correlations are hitting new highs — this time on the way up rather than on the way down. He’s not worried, though:

It is now well documented that asset class correlations tend to one during times of economic and financial market distress. 2008 was nothing if not characterized by economic distress and financial market instability. Indeed as we exited 2008 an economic depression and the end of the global financial system were at the top of the minds of investors and political leaders alike.

In that sort of environment correlations are, and should be, an afterthought. As we exited that time of extreme fear we should see a bounce back in the prices of risky assets. So the high correlations we are seeing are a natural result of the steep fall and equally steep ascent in investor sentiment and market prices.

If Tadas is right — and he might well be — then I think there are two further implications. Firstly, anybody looking at upwardly-sloping markets and seeing “green shoots” is deluding themselves: all we’re seeing here is an artifact of the financial crisis and the market craziness associated with it.

More generally, the high correlations in the market right now do seem to indicate that we’re emphatically not out of the crisis yet. And trying to apply an implicit efficient-markets hypothesis these days — the markets are going up so there must be some good news — is ludicrous: it’s precisely at times like this, when correlations are at all-time highs, that the EMH breaks down entirely.

Most of the time I trust the markets at least a little bit, not because they’re particularly reliable, but just because they’re less unreliable than anything else. High correlations are a signal to start treating market signals with extreme prejudice. Which is one reason why I’m still so bearish.