Thursday, July 3, 2008

Moody's to check on accuracy of all computerized rating models

(FT) Moody's is moving to re-examine the accuracy of all its computer models and place them under a centralised monitoring system after it formally acknowledged that a glitch had appeared in one such mathematical model used to rate complex products.

Moody's will also introduce a standardised protocol for fixing computer errors in the future - rather than letting individual units deal with problems, as at present, officials say.

The moves come after the agency admitted on Tuesday that a "coding bug" had led the agency to incorrectly award top notch triple-A ratings to about $1bn of complex instruments known as constant proportion debt obligations (CPDOs) in 2006. This bug was first revealed in a Financial Times investigation in May.

Moody's yesterday said that it hoped that the planned changes to its model surveillance should now reassure investors and stressed that it believed that the problem with the CPDO bug was an isolated incident. However, documents seen by the FT also indicate that in the past three years about three dozen additional computer bugs have also occurred in a model known as CDOROM, which forms the core of the systems used to rate most structured products.

Moody's officials yesterday said that many of these bugs had occurred in the early, development version of the CDOROM, before it was used to rate deals. In only eight of these cases had bugs had any "theoretical" potential to affect ratings, it added.

However, it has re-examined these incidents and concluded that none of the bugs did distort the ratings. "We reviewed the [bugs] and only eight of them could have had a potential impact on ratings. But in reality, none of them actually had an impact on rating," Richard Cantor, chief credit officer of Moody's, said.

"We have no reason to believe that our record [with computer bugs] is unusual," Mr Cantor said, noting that bugs often appear in software development throughout the computing world. "We do recognise though that we need to improve confidence around models and we take the complexity issue very seriously."

Jonathan Polansky, another Moody's official, said: "We have introduced a centralised model verification and validation process."

Separately, S&P told the FT it had now created an internal committee to oversee how the agency uses complex computer models. A separate internal investigation into its models has not uncovered any evidence that bugs have distorted ratings.

The moves come as policymakers are pushing for the largest credit rating agencies - Moody's, Standard and Poor's and Fitch - to take steps to enhance the credibility of their ratings process.

No comments: