by Tanta on 4/01/2008 08:12:00 AM

Our friend PJ at Housing Wire had a nice post up the other day about Michael Perry, CEO of IndyMac, and his startling announcement that IndyMac would be changing the way it collected and presented statistics on delinquent and defaulted loans. Of course I'd say it's a nice post; I contributed around two cents' worth to an earlier draft of PJ's. Self-serving motives of my own aside, though, PJ is making a very important point about how what we choose to measure--and how we define our measurements--influences what we perceive to be the risks of mortgage lending:

But if we’ve learned anything in this credit mess, it’s that all prepayments are not created equal — and that prepayments aren’t the only reason loans in a portfolio will run off.

First off, there are prepayments that are voluntary, and those that aren’t. Think of it this way: a borrower that would have defaulted in 2006 refis into a new loan in 2006 and now defaults in 2008. That’s very different sort of prepayment than a creditworthy borrower deciding to refinance because they simply want a lower payment. The real problem with the 2006 and 2007 vintages, at the core, isn’t prepayments per se; it’s that the game of musical chairs finally stopped for those borrowers whose previous defaults had essentially been “revintaged.”

So the 2003-5 vintages end up looking great from a credit perspective, even if prepayment velocity is off the charts; analysts start making complex models that only look at the effect of prepayments in whatever static pool they’ve got, and everyone declares credit risk mostly irrelevant. In contrast, the 2006-7 vintages look horrible from a credit perspective, prepayments slow and become much more volatile, Wall Street takes a look at its models and realizes some important data was missing — and, of course, lender CEOs have to pen very public explanations explaining that prepayments are “screwing everything up.”

In my view--which I laid out a bit in our March newsletter, as you subscribers will know--the prepayment picture is also muddled if you use a very narrow definition of "default." If any loan that pays in full is treated as a simple "prepayment" unless it pays off via foreclosure or you actually took a principal loss on it (that is, it "settled for less" rather than "paying in full"), then you completely miss the problem of "revintaging" as well as missing the signs of the stress level on a pool or vintage or book of loans. That's because you treat a loan that refinances while it is delinquent, or a home that sells while the loan is delinquent, as the "same thing" as a non-distressed refi or sale. If you use a more sophisticated measure of "default" that many investors do use--one that counts loans on which you didn't take a loss, but that paid off out of a prior delinquency status--then you don't get so badly fooled by the "musical vintage" problem, because you can see it coming.

Not that taking the perspective of an investor in static pools of mortgages is always helpful: that does tend to lead to the mindset that a "prepayment" means a loan "goes away" and no longer needs to concern us. What distinguishes a "static pool" like an MBS from a dynamic book of business is that in the former, no new loans are ever added. MBS "run off" by definition. Newly-issued MBS will have new loans in them that were originated as refinances, but because we're now talking about a "different deal," there is no conceptual encouragement to see these "new" loans as the prepayments from an older pool. Even the new purchase-money loans in a new pool may represent a property that "defaulted" from an older pool.

Yet investors seem to be remarkably tolerant of a situation in which very little, if any, attention is paid to where these loans in these new pools came from. It's sort of credit risk as Groundhog Day: each pool issue is new again. Borrowers and properties have no history. Prepayment analysis is always forward-looking--attempting to model the future of this pool--rather than retrospective--attempting to account for how prepayments--voluntary and involuntary, distressed and non-distressed--generated the pool we're looking at today which has not yet experienced a "prepayment."

Perhaps a concrete example will help. Shnaps directed my attention to this one in yesterday's Chicago Tribune, largely because the example given doesn't make any sense. So it's an imperfect example; I'm going to have to "make up" a couple of details in order to illustrate my point. You may reflect on why we so often seem to have to do that when reading the newspapers. I'm after other fish to fry this morning:

Janice Lee fears she will lose her 1,400-square-foot Wilmette home next month.

Lee, a former pharmaceutical representative from Minneapolis who owns Chinoiserie restaurant in Wilmette, found herself heading for trouble after she was diagnosed with lymphoma in 2003. To keep pace with her medical bills, Lee sought a $70,000 equity line on her home in 2004. Two years later, she sought a second line.

Nearly half of her $130,000 loan, or $60,000, went toward her mortgage and property taxes. But that pushed her monthly payments to $4,000 from $2,500 in two years.

In January 2007, she refinanced, pushing her monthly payments to more than $6,000, she said. She missed her first payment last March and received a foreclosure notice in June.

We do not know when Ms. Lee bought that home, or even if she borrowed money to buy it, although we have to suspect that she already had a first-lien mortgage on this loan when the HELOC series began. Otherwise we can make no sense of the payments indicated (which have to be combined first and second lien payments, or else they're payday loans.)

For the sake of example, then, I'll make up the idea that Ms. Lee bought the house in 2002 with a first-lien purchase-money loan. Why 2002? Well, cognoscenti of matters vintage will know that 2002 was once considered one of the cruddiest mortgage vintages ever to disgrace the earth. Heh. After 2005-2008, of course, old 2002 makes us all nostalgic for the "good old days." But that's kind of my point in building out this example.

So we have, in Ms. Lee's case, the following appearances in the following vintages:

2002: A new purchase-money first-lien loan
2004: A new cash-out HELOC second-lien loan
2006: A new cash-out HELOC second-lien loan that pays off the loan in the 2004 vintage
2007: A new cash-out first-lien loan (I think) that pays off both the 2002 loan and the 2006 loan and that is in FC. It was also, you note, an EPD (early payment delinquency), since it seems to have missed either its first or its second payment and was in FC by payment 5 or 6.

If you do a certain kind of simple-minded "vintage analysis" of the kind PJ is complaining about, you would get the following:

2002: A good vintage, since the loan never defaulted and paid in full
2004: A good vintage, for the same reason
2006: Ditto
2007: A very bad vintage

But what, really, about those earlier vintages was so "good"? Were these "good loans," or did we get lucky by having another lender around willing to "revintage" the loans via refinance? From hindsight, the lender in the earlier vintages looks like it got lucky, because someone else was holding this bag in 2007 when the music finally stopped (it's obligatory to mix metaphors in this context). At the time, of course, they might well have been complaining bitterly about prepayments erasing their yield on those pools (or their servicing income). In fact, they might have been so bitter about it that they developed these noxious "prepayment penalty" things to keep those apparently "good" loans in place. Yeah, that looks like a good idea now, doesn't it?

The Tribune article, of course, doesn't tell us whether Ms. Lee was ever delinquent on those earlier loans. I suspect she was, since it looks like she was paying real subprime interest rates on at least the last two, and that might well have been because her prior mortgage history wasn't good. If that's true, then the earlier vintages really dodged a bullet here: they escaped a loss on the loan only because a greater fool stepped in to refinance it.

Thus, as PJ notes, the problem with our most recent vintages: the greater fools got run over by a truck, and so loans aren't "moving" any longer. They stay where they are until they finally fail. It will undoubtedly take a long time until we get another vintage as ugly as 2007-2008, but that's not just because (we hope) it will take a while for memories to fail and lending standards to become as stupid as they have been. It's because the lack of an "exit" means that those vintages will be forced to "show" the real defaults.

It is also important to really notice the implications of a point PJ makes here: if you look not at individual pools of loans but at the entire outstanding "book" of subprime and Alt-A loans in the aggregate, you are going to see "rising" delinquency numbers even if "nothing gets worse" than it already is. That is because, until further notice, no or extremely few new subprime or Alt-A loans are being made. The whole "book" is in "run-off" mode. That means, if you use a "current balance" to calculate delinquency and default, the "current balance" just keeps getting smaller and smaller, because no new balances are added to it. The average loan age just keeps getting older, for the same reason. In other words, we really do, for once, have a "prepayment" situation in which liquidated loans just do "go away." You can think of lender REO inventory as exactly that: the old loan for the old owner "went away," but since there's no new buyer wanting a new loan, there is no new "loan vintage," just a nasty REO inventory in a kind of "limbo."

What Perry of IndyMac is up to, of course, is deciding to quit reporting "raw delinquencies" on current balances right at the time when that kind of statistic is going to just keep looking worse for a long time, and substituting an alternative kind of reporting that will look better. That doesn't mean that the alternative reporting is "false." It means, in the most generous case, that we're looking at the part of the cup that's half-full. But switching measurements (and universes of loans to be measured) at the beginning of the unwind is going to play havoc with our ability to understand history. Those of us not looking forward to being doomed to repeat it do care about that.