Moody's Investors Service announced that it has modified its approach for rating synthetic collateralized debt obligations, moving to Monte Carlo simulation-based models from binomial approaches. In many circles, particularly Europe, the change is a significant one.
Dubbed "correlation-intensive structures," the list of securities affected includes static synthetic CDOs-of-CDOs, CDOs of equity default swaps and CDOs with long and short swaps. Since about 2002, issuance of these securities has boomed, leading all of the ratings agencies to look more closely at how the risks are evaluated. For Moody's, the move to Monte Carlo was a natural one, said Bill May, a managing director at Moody's.
"When a market grows as rapidly as the static synthetic market has, you look at that and think about the best ways to assess credit risk. It naturally focuses your attention," said May. "Tomorrow, if CFOs [collateralized fund obligations] went through the roof we would look more closely there as well."
Recently published statistics from Banc of America Securities showed a 22% growth rate in synthetic issuance over the year and an eye-popping 71% growth in synthetic static arbitrage deals alone. BofA recorded $216.04 billion issued in these securities in 2004 as of May 21, compared with $126.54 billion for the same time frame in 2003.
"Especially with the advent of the indices in the market... these have really caused a ballooning in the number of deals that are being done," May said, referring to IBoxx and Trac-X, the two major traded synthetic indices, which merged in April.
As Rodanthy Tzani, a vice president at Moody's, explained, the switch in methodologies was not due to drawbacks of the binomial approach per se, but because "we can do something better given the recent shift in market conditions and the emergence of new products," she said. "[Using a Monte Carlo approach], we can look at each specific name individually, and this is a more exact way of rating these transactions. Synthetic transactions are quite sensitive to subordination levels, for example," Tzani added.
The market will see a new set of assumptions for inter- and intra-industry asset correlations plugged into the Monte Carlo simulation used to calculate the expected loss of a static synthetic CDO. In addition, an implicit inter-industry correlation of zero was incorporated for corporate pools, but now Moody's will assume a positive number, while intra-industry correlation will be reduced slightly. Finally, a beta distribution for recovery rates, using the same parameters already found in Moody's Recovery Rate Toolkit, will be implemented.
Perhaps the most visible change, though, is the fact that Moody's is distributing the model to issuers, allowing a better approximation for what the deal's rating might be once Moody's analysts plug it into their models.
"This allows them to structure the deal and, to some extent, market their deal without taking up a lot of rating agency time. It gives a tool to approximate what our rating will probably be," May added.
The effort to revise the methodology stemmed largely from the growth in CDOs-of-CDO transactions.
"The CDO squared [transactions] have the peculiarity that correlations between tranches of CDOs in the portfolio are quite high. And some of this risk cannot be assessed by using traditional methods," Tzani said. "Monte Carlo takes into account individual credits, simulates underlying names, calculates losses of the first-level CDOs, correctly reproducing the correlations between those CDO tranches, and finally, gives the expected losses of the rated tranche," she explained.
Copyright 2004 Thomson Media Inc. All Rights Reserved.