walkitout (walkitout) wrote,

Fun with RMBS, or, that whole mortgage mess

I read a little thing on Calculated Risk about Fitch updating their model for rating RMBSs (securitized mortgages, which no one is buying now, but never mind that for the moment).


Now, I have a lot of trust and respect for these bloggers, but 25% drop nationally in real terms? From now? Let me look at your citations. I mean, it isn't that I don't trust you; it's that I want a bit more detail, particularly when the range is from 7% appreciation in San Antonio to 47% depreciation in San Diego.

I had to register over at Fitch, but they didn't ask me for any money and I didn't lie to them and that got me an 11 page .pdf report released today. I haven't made it past the first page, NOT because I don't understand it (I am sorry to say, I actually do, more or less) but because it is really freaking me the fuck out. But first, a little background. (ETA: the remainder of this post is essentially me MST3King the Fitch report).

Let's say you want to buy a house. You go to a mortgage broker. You fill out some paperwork. Some amount of checking is done on that paperwork ("underwriting") and signed off on by the broker, who then hands it off to a bank doing a wholesale mortgage operation. Assuming one still exists. Which it probably doesn't. But let's pretend it's 2007 and all this stuff was still happening. The wholesale mortgage operation, which has a relationship with the broker, gives the broker some money for the broker's trouble along with the money for the mortgage. The wholesale mortgage operation stuffs your loan -- along with dozens, hundreds or wtf other loans -- into a package (an RMBS) and sells it to investors, like, little old ladies, other banks in need of reserve capital, hedge funds and so forth. The package has some rules (thou shalt not modify more than some small percent of these loans) and some assertions (only some small percent of these loans will have late payments/default/go into foreclosure/be paid off early/etc.) associated with them which will be key points in marketing them.

Because the little old ladies (and municipalities, non-profit organizations, banks, etc.) who are buying these things would like some assurance that they are what they say they are and will do what they say they will do (pay a certain amount in interest, generally), the loans are first vetted by ratings agencies, who run the contents of the box (this is all virtual, mind you, altho in theory there are literally boxes of paperwork following the bits around, in practice, the paperwork often gets lost and clever lawyers in foreclosures can stall things indefinitely forcing people to find the paperwork in the box which got lost) through a MODEL which tells them how likely it is that this thing is what it says it is, which is to say, is it a turd or is it a golden turd or is it gold.

We all know the GIGO thing: Garbage In, Garbage Out.

Fitch has taken the opportunity presented by a completely non-existent market for RMBSs to rework its model. R. notes that if one of the ratings agencies has a stricter model than the rest, they will not get any business. Well, with no business in sight, that's not a very compelling argument. There might be some regulators or, at least, some AGs breathing down their necks as well. Fitch, therefore, is motivated to Improve Their Model. What improvements have they decided to make?

In no particular order, and not a complete list:

They have frequency of foreclosure numbers. Fine. They've updated them. Yay. NOW, they treat a loan differently if it's an 80/20 with no second mortgage vs if it's an 80/20 with a second mortgage. This one more or less stopped my heart right there. Yes, Virginia, apparently previously there was _no impact_ on the frequency of foreclosure expectation between an 80/20 where I actually showed up with the 20% down cash at closing, vs. I got some other bank to loan me the 20% at the closing. Well, no _wonder_ everyone was doing it. I now start to believe in the 25% drop in the market, real terms, starting now.

Also updated: there's now a bit of a negative associated with "loans underwritten to a program where income or assets are not verified". Heart had restarted in time to stop again. So, Fitch (and, presumably, S&P and Moody's) could not give a flying leap previously whether or not anyone actually did the underwriting, in terms of producing a frequency of foreclosure prediction. NOW they care. Horse gone. Barn door locked. I'm wondering, is that 25% drop overly optimistic? Might be -- that's Fitch's base case after all.

These two changes combined change default expectations by 4X. Ya think?

I would also like to note that Fitch expressed surprise at the fraud rates they uncovered back in November of 2007. Which is apparently the first time it occurred to them to look.

I asked R., if you were teaching a class in simulations, and you had a student slap together a model for housing prices and they only included the case in which housing went up in value, but totally failed to include the possibility that housing might decline, what kind of grade would you give that student? He was really nice. He said they wouldn't be getting an A.

I'd fucking fail their ass. I don't care _what_ their parents donated to the university.

In this context, I'd like to note that NOW Fitch accounts for change in home price between loan origination and loan current age to "account for the higher risk of default and loss in a declining price environment".

Let me be _UTTERLY CLEAR_. Kudos to Fitch for making much needed changes to their model and for announcing them. Steps like these are crucial to restoring investor confidence, which in turn is absolutely necessary for restarting the secondary loan market, which in turn is necessary if we aren't going to spend the next decade or so debating whether or not this is as bad as the Great Depression or worse. Yay.

If/when RMBSs become marketable again, they won't be marketing brand new ones (which, honestly, was pretty weird anyway). They'll be marketing "seasoned" loans, which is the way things used to be done. You went to your bank. You filled out paperwork. They checked the paperwork and asked for more. They loaned you the money. You made payments for a few years. THEN they sold your loan and you got to play the where-the-hell-do-I-send-the-check-this-month game. Fitch is making some of these changes in support of its efforts to assess "seasoned" loans, which tend to be more predictable (after all, if you managed to make payments for a couple of years, that's a pretty good sign, compared to some liar's paper that just crossed the desk and hasn't even had a single payment on it yet).

Unfortunately, those "seasoned" loans apparently will be including some of the loan mods that lenders have been pressured into making over the last few months. Which, I might note, in general do _NOT_ include writing down principal, but just roll costs and etc. into something that is maybe fixed for some period of time and maybe has a lower interest rate for some period of time and so forth. Think of them as a form of denial. Fitch is much nicer than me; they call them reperforming loans (holy shit! that is the most _ridiculous_ thing I've heard. All day, anyway), or "scratch-and-dent" loans (more like it, only I'd call it totaled with a whitewashed title, but hey. I'm mean.).

Just a little side note here: Fitch has just introduced metropolitan area indexes. They used to figure frequency of foreclosure adjustments etc. on a _state level_ basis. Words fail. I mean, we _are_ _in fact_ discussing residential real estate. And there are two things that you hear so often about real estate that after a while you just want to slap the next person who says them (if you're me, even if I say it to myself). One is location, location, location and the other one is all real estate is local. _State Level_. Yeah. Cause there's definitely a lot in common between, say, Renton and Yakima. Berlin and Hollis. Lee and Boston. Temecula and Fremont. Just as a for instance. They have switched over to what sounds like a Case-Schiller metro-area approach. _That_ seems reasonable. They spend a paragraph or so on why. I think you can work that one out on your own. They've also added a national adjustment, and they've hired some other people to provide them with quarterly projections for all these little adjusters, which are intended to represent macroeconomic trends.

If you're thinking, hey, that sounds like a GSO to me! you clearly hang out with the same people I used to. A GSO, for the rest of you, is a Graduate Student Oracle. You need a constant to make sense of your data. You have no idea what it is. You ask a graduate student to gin up a number that makes the whole thing kinda work. Graduate Student Oracle.

There's more. A lot more. They're kinda pro-rating the loss severity credit of having mortgage insurance, since mortgage insurers aren't paying out as much/as reliably (usually because of fraud and "linked to first payment default". See, I shouldn't read this shit. Loans moved so fast they had problems with mortgage insurance denials on loans where not one single payment was ever made. Ever.). There's also a discount to allow for macroeconomic stress impacting the insurers ability to pay out.

As for that 25% drop from now? That's fallout from their GSO, er, 5 year home price forecast from University Financial Associates, LLC (UFA).

"By weighting the outstanding loan balance by state from data obtained
from Loan Performance and mapping the results to the home price forecast provided by
UFA, Fitch projects home prices to decline by an average of 25% (in real terms) at the
national level over the next five years (from second-quarter 2008). This represents
Fitch’s base case home price decline assumption."

So. Fitch really did say it, in their out loud voice. The question is, where did UFA pull this out of?

For that matter, who is UFA?


Their title includes: Subprime risk management consulting and software development

Reviewing the various things they have for hire, it's reasonable to conclude that Fitch bought a lot of their recent upgrades from UFA, one way or another. Given what I'm looking at, I have to conclude that just sticking the Fitch name on something from UFA (particularly in the loss of reputation all the raters have suffered from recently) doesn't really make a helluva lot of difference one way or the other. Maybe I'll send e-mail to one of those people I know who used to work for MILA and ask them what they think of UFA.
  • Post a new comment


    default userpic

    Your reply will be screened

    Your IP address will be recorded 

    When you submit the form an invisible reCAPTCHA check will be performed.
    You must follow the Privacy Policy and Google Terms of use.