Mike, you make some good points. However, in closing my post I raised a question that should be given some consideration -- what is it that's being criticized?
-------
An HSBC exec is quoted in the article as saying that FICO scores were "less effective or ineffective" in predicting behavior during a period of aggressive lending and low interest rates.
I doubt there was any question in HSBC's mind that they would suffer a higher default rate in their subprime portfolio than with higher credit borrowers. But the HSBC statement suggests that they feel the FICO product in question didn't appropriate identify that risk.
We don't know what FICO asserted for this product. We do know that when FICO profiles our personal scores that it cites very specific default risk for each credit score band. I don't think it's unreasonable that thie FICO product did as well for this subset of subprime borrowers -- quite possibly in a very well defined manner.
HSBC may have relied upon this data to determine what mortgage risks they would underwrite and at what rates - FICO offering up a fairly specific indicator of expected default cost under their model. If the model proved to be inaccurate and a higher than indicated default rate occurred (taking into account economic conditions), then there's room for lenders to hold FICO at fault, at least in part, for the unexpected poorer loan performance.
------
The article gives good indication that this perceived ineffectiveness in predicting defaults is the crux of the matter. Note the following statement from Vantage:
"Absolutely, it helps us," said Barrett Burns, VantageScore's chief executive. "It offers us a great opportunity for lenders to look at VantageScore. Since our score is more accurate and predictive, there would have been improvement" in mortgage defaults had the industry used VantageScore instead of FICO, he said."
The gist is that the Vantage product is asserted to more accurately predict default rates for these borrowers then FICO. To the extent that Vantage should more accurately identify a stronger default risk vs. FICO, lenders would reign in lending (using the same underwriting standards) and, thus, experience lower defaults -- or, at least would price loans more accurately and thus not suffer the adverse profit consequence of the higher default rate.
------
What's problematic here is that we don't have any data whatsoever from which to assess whether the FICO product was inaccurate in predicting degree of default risk. Short of that, we're in an inadequate position to make a reliable call on the situation.
Still, the statement from Fitch Ratings in the latter part of the article gives good credance that the FICO product isn't likely at fault. Fitch argues that lenders, anxious to assume these loans, weakened underwriting procedure by, amont other things, requiring less income documentation. The consequence of inadequate income doc is that if borrowers successfully misreprented their income as stronger that the actual case, the lenders put themselves in the position of likely overextending loan amounts.
In other words, FICO default predictions are based upon a given underwriting process (those standards prevalent during the period over which data was compiled and analyzed). If lenders should then become more aggressive in pursuing the subprime market (presumably because they have new confidence in possessing a new tool that will more accurate predict defaults) and make the application process less stringent in the process (e.g. lower loan doc reqs), then actual default rates will necessarily vary from those predicted by the FICO model.
Consider this analogy: We're told in our myFICO score report analysis that 14% of borrowers with scores between 650 and 699 will "get into serious credit trouble". This assumes that lenders continue existing lending practices. However, if lenders began assigning much higher credit limits for applicants in this credit band, or simply approve a higher percentage of applicants, there's no question that default rates would rise. The FICO score predictor would be meaningless in light of a change in lending practice.
------
We don't have sufficient data to make a definitive call on the questions raised here; we can only make some reasoned guesses, at best.
Personally, I find it much more probable that lenders have inapporpriately applied the FICO data in making underwriting decisions than that the FICO model predictions were inaccurate.
But I'd be deluded if I thought there wasn't good room for things to be just the other way.
- Harry