by Tanta on 8/18/2007 12:56:00 PM
Saturday, August 18, 2007
What was interesting about the Lehman analysis is that it looked quite squarely at the possibility of what it calls the “moral hazard borrower,” or some set of borrowers ending up getting a mod when they wouldn’t have defaulted (or could have gotten a market-rate refi) because the servicer’s “targeting” was too wide. They conclude, actually, that with careful enough cost-to-trust analysis of the terms offered on all mods, and limiting mods to borrowers who have already become delinquent, the “moral hazard borrower” problem isn’t likely to cause noticeable losses.
If I get permission from Lehman I’ll post some more of the analysis. Until then I think I can get away with this snippet regarding the methodology of their analysis, which I bring up for discussion purposes:
In our scenario, we assumed that the proportion of the borrower pool in each of these groups [current borrowers, those who can be cured with a mod, those who cannot be cured, the “moral hazard borrower”] depends on the amount of rate or payment reduction. Because we did not have data on borrower responsiveness to loan modifications, we extrapolated the sensitivity of defaults from observed response of subprime ARMs to payment shocks. A 9% payment shock on subprime ARMs has historically caused about a 20% increase in the credit default rate (CDR). The experience from payment shocks is not entirely applicable to the loan modification scenario as subprime ARM borrowers who choose to stay on with their mortgage post-reset are adversely self-selected. The higher default rates from this self-selection cannot be directly distinguished from the economic impact of the payment shock. However, given the lack of data on the likely impact of loan modification, we are using the response to payment shocks as a proxy for the impact of loan modifications.
What might it mean to say that “borrowers who choose to stay on with their mortgage post-reset are adversely self-selected”?
First, we must assume that a choice is a choice. We therefore assume that there are alternatives, such as a refinance at a rate/payment lower than the reset rate/payment that these borrowers could qualify for but choose not to, or that the property could be sold without financial hardship to the borrower, or the borrower could simply mail in the keys.
Any borrower who “chooses” to keep a loan with a 6.50% margin that resets every six months instead of refinancing or selling at break-even or walking away is, therefore, presumed to be:
4. Making plans to scarper
5. Running a tax dodge
6. So traumatized by the original experience with a loan broker that he or she is unable to contemplate going through that again even if it means starving
7. A couple of tranches short of a full six-pack, if you know what we mean.
The inescapable conclusion (you might want to sit down for this, it’s stunning) is that it is very difficult to model the behavior of this group with the usual variables like “in the money rate incentives” or “moral hazard” or “damage to credit rating” or “pupil dilation in presence of bright lights.”
At this point, we merely pause to recognize the nature of what this is saying about the subprime 2/28 and 3/27 ARM: it was never intended to be a 30-year loan. It was always a bridge loan pretending to be a 30-year loan. It cannot be modeled as a 30-year loan. But hey! It’s a great product for achieving stable homeownership goals!
In any event, as of today we’re well past that point where “choice” and “self-selection” are the operative mechanisms. In the current environment, we have:
1. Little available refinance money (lenders are not lending)
2. Little available refinance incentive (refi rates are high when they are available)
3. Little available refinance flexibility on high LTVs (the “add-on” cost for a high-LTV loan is no longer artificially lowered by “nontraditional loan products” manipulating the payment)
4. Little opportunity to sell for at least the loan amount plus transaction costs
This means, as far as I’m concerned, that it is quite likely that the universe of “post-reset borrowers” is no longer adversely self-selected. It may well be adversely selected: “little opportunity” to sell or refi does not mean “no opportunity,” and so the very highest-quality borrowers and the properties in the healthiest RE markets will opt out of the pool. But your remaining pool is not the “classic” self-selected group any longer.
This is why workout options like mods start to make sense: the pool of defaulting borrowers is no longer exclusively the group of people for whom little can be done; the pool includes people for whom the credit crunch removed what could have been a viable option. In a credit crunch, the model that assumed “adverse self-selection” no longer works reliably.
Beyond The Great Modification Controversy, I think it’s worthwhile to return to this question of how much “historical” data we ever had to justify all the risk we put into the system during the great lending bubble. A lot of people cheerfully made these 2/28s because they based their “stress test scenarios” on past episodes of economic or housing market distress in which different products were offered to borrowers (either fixed rate loans or straight “bullet” ARMs that don’t have this initial teaser/IO fixed period/prepayment penalty combo). The only excuse for this, which is now becoming explicit, is that we just counted on easy refi money and endless HPA to take care of the problem. I know that’s not news to the Calculated Risk crowd, but it still seems to be news to these CEOs (ahem) who stand up and say “no one saw this coming” and “we didn’t lend on appraised values.”