The latest revisions to the model will impact Fitch’s expected loss assumptions and credit enhancement levels for RMBS, the company said; what’s perhaps most surprising about the update, however, is that it suggests just how rudimentary some of the ratings process really has been.
For example, among the changes in ResiLogic 2.0 are the introduction of MSA and national macroeconomic risk multipliers, which will allow the model to fine-tune its estimates of both frequency of foreclosure and loss severity around a series of 25 key regional multipliers, as well as a global risk multiplier intended to capture the overall macroeconomic state. (The previous ResiLogic model used neither and relied only on state-level risk multipliers to adjust risk.)
Perhaps this is because I’m a former econometrician, which means I’m still very much a geek, but the apparent fact that the prior ratings model used to assign ratings to a given RMBS deal couldn’t already account explicitly for “stressed macroeconomic trends” and adjust the risk of a deal accordingly is eye-opening.
The additional fact that the rating model apparently only used state-level risk multipliers updated quarterly is stunning; the new model will use quarterly data, as well. Call me too new-school for my own good, but the data needed to compile MSA, state, and national level risk multipliers is available monthly from a wide range of data providers; I know, because I’m lucky enough to see the data each month. If we’re ever going to be talking about a multi-trillion dollar private-party securitization market again, I’d expect investors to demand that their deals price risk in using the most recent data available.
An ever-so-slightly outdated risk factor can mean the difference between a profitable residual and one that is quickly wiped out, after all.
This isn’t meant as a shot at Fitch, which has consistently been among the most conservative of the big three credit rating agencies both before and during the mortgage crisis; they’ll probably remain the most conservative when this mess is eventually put in the rear-view mirror, too.
But I’ve heard from more than a few market participants, the kinds of in-the-trenches people that would never get permission from their employer to see their name in print, that the opaque ratings models used by the big three — a “competitive advantage” that all three firms work hard to protect, hence the opacity — also allow the agencies to cover up for models that have some very rudimentary failings in modeling techniques.
The risk factors discussed above would appear to be one example of this, and while it’s good to see Fitch now making an attempt to correct for this, the move to include the variables into the model now is really no different than closing the barn door after the animals have long since run out of their stalls.
And it’s only one example of how ratings models missed clear risk factors that might have forced greater credit enhancement, but might have also protected investors to a larger degree: Fitch said its Resilogic 2.0 model will up penalties for loans written in no-doc or low-doc scenarios, as well as for loans with a piggyback second lien at origination.
There was a time, two or three years ago, when I would have been surprised by these sort of revelations, having held the agencies and their analysts in such high esteem; their models fueled a market of almost unfathomable size, and I’d admired that. Now, however, just like the market they helped build, I’m realizing that much of it might have been nothing more than a house of cards.
Which means fixing this market is going to take time.