Tuesday, August 07, 2007
As The Stone Rolls Down The Hill
--Greater emphasis on regional economic risk. The ResiLogic model will give greater weight to the University Financial Associates (UFA) default multiplier component of the model. The UFA multipliers, which are updated quarterly, raise or lower the ResiLogic expected default probability as a function of a state-level forecast of default risk. This revision affects all mortgage products. Expected default rates will increase on average approximately 20%, depending on the geographic distribution of the mortgaged properties.Tanta at Calculated Risk hasn't stopped hyperventilating about the stupidity of some of the assumptions they were using. According to the statement, Fitch has now decided to use back-end DTI instead of front-end. I went to the ResiLogic page and downloaded the documents. It wasn't clear to me what ratio they had been using, because this is the passage:
--Increased default expectations for short-term and hybrid adjustable-rate mortgage (ARM) products. Expected default rate multipliers for 2/28 ARMs are increased by 22% and by 12% for 3/27 ARMs. Expected default rate multipliers for 5/25 ARMs are raised by less than 5%. Expected default rates for ARMS with adjustment terms below two years are increased by 30%. Option ARMs are subject to an 8% higher multiplier.
--The introduction of a new risk category for borrower income and asset documentation. ResiLogic now categorizes certain loans as 'Low' documentation, in addition to the existing 'Full', 'Reduced' and 'None' categories. Expected default rates for 'Low' documentation loans are 43% greater than for 'Full' documentation loans.
--Default expectations will be determined through analysis of 'back-end' debt-to-income (DTI) ratios, rather than 'front-end' DTI ratios as had been implemented in ResiLogic. If a debt-to-income ratio is not provided for a subprime loan, the default DTI is assumed to be 50%. The impact of this revision is generally minor, although subprime data files lacking in DTI information will be noticeably affected.
--The removal of benefits for short-term loan seasoning. The ResiLogic model reduces default expectations for performing loans as a function of loan age at the time of securitization. This feature of ResiLogic has been removed.
Debt-to-Income Ratios (DTIs): Default rates were slightly correlated to increases in front-end DTIs, as shown in the chart above. A front-end DTI is the ratio of monthly installment debt to monthly gross income (the mortgage payment, taxes, and insurance are included in the back-end DTI). For prime and Alt-A loans, a 5% increase in the DTI yielded roughly a 3% increase in default rates.Lord only knows what they were really using. Normally front end would be mortgage payment to income, and back-end would be total payments to income. Anyone who really believes that FICO scores are more predictive of default than total debt payments/income is an idiot, that's all I can say. Do you really think that someone scheduled to make payments of 70% of their monthly gross income every month is likely to perform long term? Tanta isn't wrong - this reads like amateur hour. Anyway, the original document sounds like they were using all debt except mortgage-related, which can't be true.
For subprime loans, there was slightly less sensitivity to default risk when DTIs rose. A 5% increase in DTI yielded a 2.5% increase in default risk. This lower sensitivity is also reflective of the very high default rates of the subprime sector and the vulnerability of the subprime borrower to external risk factors
However, the fascinating thing is that according to their update, they are still not using a very strong model. They have increased the state-level modelling, but it is not good enough by a long shot.
You can see why the market is twitching so severely - they don't really have an idea of what to expect, and the gatekeepers are blushing and saying "lalala - oops - who'da thunk?"
In other news, Compucredit's second quarter earnings were released last week. They took a net loss of 23 cents a share. Their net interest margin has been bouncing around quite a bit as they try to balance income against defaults:
CompuCredit's net interest margin was 18.9 percent in the second quarter of 2007, as compared to 23.7 percent for the second quarter of 2006 and 17.9 percent in the first quarter of 2007. The adjusted charge-off rate was 9.2 percent in the second quarter of 2007, as compared to 8.3 percent for the second quarter of 2006 and 12.5 percent in the previous quarter. As of June 30, 2007, the 60-plus day delinquency rate was 13.2 percent, as compared to 11.5 percent as of June 30, 2006 and 12.8 percent as of March 31, 2007.Standard & Poors put 207 Alt-A classes on ratings watch negative. It's an even bet as to whether any Fed head will make another statement about "containment" to subprime. Alt-A & subprime originations are considered to have composed about 40% of all originations last year. Rapidly increasing jumbo rates are also going to put pressure on states with higher property values.
I really liked this comment on the post at Calculated Risk by Kirk Spencer as a description of the likely sequence of events leading to the peak of the credit contraction:
I think we're in the first of a three-stage slow motion whammy, and the reason it's slow is it's all tied to real estate and real estate is, at heart, a slow-motion process. (A house doesn't go through several hands a day every day -- usually.)Prime loans in quite a few areas will be facing somewhat higher foreclosures but significantly higher loss severities than normally because of this sequence of events. Of course, residential mortgages are only one aspect of easy credit that is becoming much tighter credit. There is almost no aspect of the credit market that hasn't been affected by loose standards.
Whammy one is the realization that a lot of what's holding up the market has been artificial exuberance. It's been - so far - tied to the loans made to people who couldn't afford it and the speculators. It's beginning to dawn on people that the risk-loans (below AAA bonds) are inflated, and with loss of trust comes loss of money. We're also seeing the barn-door locked with tightening standards which effectively reduces credit, but that's not hitting us yet.
No, that starts to hit us in - I'm guessing - late November or December. That's when we see that peak of resets. And all the people who are looking for refi's -- who probably could have qualified for one of the risky loans a year ago -- can't get their new loans because the risky loans aren't there for them (and they can't qualify). And they start talking foreclosure and jingle-mail.
And that leads us, in about three months (March maybe April next year) to the third and worst whammy. That jingle and foreclosure process is no longer tentative but complete, and suddenly all those banks are carrying properties (expenses) instead of loans (receivables). And they're hoping - anticipating - unloading in the selling season which just started. But nobody's buying. They're not buying for two reasons. First, a lot of people now know there is an excess of inventory. Second, a lot of people who'd like to buy anyway can't because there's no credit -- second whammy magnified. It's around this point that the bank's stated values are insufficient to support the trades they've offered - the REALLY GOOD ones, the ones thought to be AAA.
It's almost nine months out. Smart companies can (and sill) start arranging now to minimize the pain. If enough companies are smart and are willing to take pain to reduce it later, then this will not be much more than the sort of thing we see now. If, however, they all insist on waiting to the last minute... we're screwed.
Caveat -- things like this recent Fitch announcement and the BoA deal in California may accelerate the whole thing, making business attempts to mitigate too little too late, but we'll just have to see.