Credit market's correlation models break down

LONDON, Feb 21 (Reuters) - Correlation is a measure in the credit markets of investors’ fears of a few corporate defaults versus an economic catastrophe, but credit pricing is now so awry that correlation models have lost sight of reality.

Even stranger, the mathematics used to design traditional correlation models has reached the point that it no longer works, analysts say.

The problem may be theoretical, but it is having a real impact on the credit default swaps market in recent weeks, because dealers need these models to manage their portfolios.

Some banks are reworking their models and then doing trades to rehedge against their risks in line with the new models, which is adding to volatility, analysts say.

“What is happening is that you improve the model to match this new kind of market reality,” said Efrosyni Charalampidou, a credit strategist at UBS. “That is what many structured desks are trying to do.”

Correlation is based on portfolios of credits, such as collateralised debt obligations (CDOs), which have been divided into tranches based on different degrees of default risk.

The riskiest slice, known as the equity tranche, is exposed to the first few defaults from any company in the portfolio and is completely wiped out when losses reach 3 percent.

Losses then move to the next tranche up. Tranches at the top of the ladder are typically rated triple-A, because a large number of companies would have to default to affect them.

Correlation measures the relative value of tranches at the low end of the ladder, which are vulnerable to idiosyncratic risk, versus tranches at the high end, which are vulnerable only when the entire system comes crashing down.

Typical correlation prices are based on standardised tranches of the leading credit default swaps (CDS) indexes -- the Markit iTraxx in Europe and CDX in the United States.


Spreads on the highest AAA tranches of the CDX investment-grade index have widened so much that they go beyond the maximum spread levels that models can allow.

At the maximum correlation allowed, the models indicate theoretically that if one company were to default, all companies in the index would default together, UBS analysts wrote in a recent note to investors.

“Mostly the problem is in the States, but because there is one model across the world, everybody has to play around,” Charalampidou said. “People do not have one model for Europe and one model for the States. There is one model for the correlation market.”

Correlation desks are altering these models mostly by changing their assumptions for recovery rates after defaults, she said.

Abel Elizalde, a credit strategist at Bear Stearns, said the typical correlation model has three main variables -- spreads, which are determined by the market; recovery rates, typically fixed at 40 percent, based on historical averages; and a correlation measure of relative value across tranches.

As spreads rose in senior tranches more than in the underlying indexes, banks adjusted their figures for correlation, but the figure has got so high that it no longer makes sense mathematically, he said.

The one variable that can still be altered easily is the recovery rate. If assumptions for recovery rates are reduced, then fewer companies must default before the top tranches suffer losses, he said. “It makes the calibrated correlation number more reasonable.”

Analysts cite what they call technical factors to explain why correlation levels have decoupled from reality. That means it has little to do with investors’ views on whether prices of AAA tranches compensate for the risk.

According to UBS, the current implied default rate for the investment-grade iTraxx index in Europe, even with the recovery rate at zero, is still above the worst historical scenario.

“In normal circumstances, we would be at the point now where longer-term players would step in with the view that spreads are really attractive,” said Michael Hampden-Turner, a credit strategist at Citigroup.

One restraint on would-be investors, however, is that accounting rules force them to take writedowns every quarter when prices of derivatives positions fall, even if there have been no losses from defaults and they plan to hold to maturity.

“Many investors we talk to have been saying, ‘I think treble-As are really attractive right now, with spreads way too wide relative to fundamentals. But the markets are extremely volatile, and if I buy now, I could lose money to begin with, even if I make money in the long term. I am not prepared to jump in until volatility dies down’,” Hampden-Turner said.

Volatility, therefore, is turning into a vicious cycle.

“Higher volatility is effectively reducing liquidity in the market, and the lower liquidity is causing a higher impact by the trades that do get executed,” Bear’s Elizalde said. (Editing by Will Waterman)