In Depth Analysis: CalculatedRisk Newsletter on Real Estate (Ad Free) Read it here.

Saturday, May 05, 2007

MBS For UberNerds III: Credit Risk, Credit Enhancement, and Ratings

by Tanta on 5/05/2007 11:16:00 AM

Prior Installments:

MBS For UberNerds I: GSE Pass-Throughs
MBS For UberNerds II: REMICs, Dogs, Tails, and Class Warfare

Last episode, we looked at the classic multi-class mortgage-backed security, the REMIC, from the perspective of its distinctive cash-flow issues. REMICs can and do take an existing GSE-guaranteed single-class pass-through MBS, or several of them, and “tranche them up” into multiple classes with differing payment, prepayment, maturity, and yield to investor characteristics. That’s relatively easy to do, because the underlying collateral is guaranteed from the perspective of credit risk and timely payment of principal and interest. But REMICs (and REMIC-style CMOs) can most assuredly be backed by whole loans, or by single-class MBS that are not agency-guaranteed, such as Alt-A, subprime, HELOCs and HELs, scratch and dent, reperforming, seasoned (old loans), even nuclear waste (nonperforming). How do you deal with not just the cash flow but the credit risk of that?

From a loan quality perspective, the pool backing a simple GSE MBS is a bag of Raisinettes—whatever serving size you take, you get the same candy in it, with the natural variation God gave a raisin. A REMIC pool, on the other hand, is more like a bag of Bridge Mix. There can be jellies in there. There can be a lot of them, and some of those “raisins” are possibly, um, insufficiently raisin-like. As a matter of fact, some of these private-issue pools can go out of their way to add a substantial chunk of poor-quality loans to a pool, because the pool “needs” the higher interest rate those loans pay in order for its cash-flow calculations to work out (we’ll get into that below).

What is required is “credit enhancement” (CE): something that makes up for the greater likelihood of default in the underlying mortgages. No private-issue REMIC is guaranteed, the way GSE MBS are. They all have at least potential risk to the investor of loss of principal, as well as uncertain timing of payment of principal or interest. But because they are structured, they end up with tranches of varying degrees of credit risk. This allows each tranche to receive its own rating, and the rating of the tranche can be much higher than the rating of the underlying mortgages. As this fact has caused more confusion than nearly anything else I’ve seen lately, we will go into some detail.

First, do remember that individual loans can be “credit enhanced”; that’s what mortgage insurance is. Some individual loans are credit enhanced by, basically, being tranched themselves: that’s what an 80/20 deal is. (Which is why CE on a HELOC pool is a whole different bowl of chocolates from a first-lien mortgage pool.) Whole pools can also be enhanced with MI: that’s a pool-level policy.

There can also be other kinds of insurance on the security, such as a letter of credit or surety bond, although these forms of “external” CE are less and less common. Part of the reason for that is the rating issue: with an externally-enhanced security, the rating of the security can be no better than the rating of the guarantor. (Remember that GSE MBS get their AAA rating not because each individual loan is AAA—mortgages to just plain folks like us don’t get better than “A” ratings—but because the guarantor, the GSE, has an implicit AAA rating.) You don’t find enough AAA-rated banks writing letters of credit against subprime pools for that kind of credit enhancement to work out. And the best of the MIs are AA or A; that’s not enough, by itself, to get you an AAA rating on any part of your security.

What you get, then, is generally “internal” credit enhancement. Remember the “sequential pay” idea? We went through that from a perspective of cash-flow and time-to-maturity, but it has, actually, an imbedded or implicit credit enhancement function. Think of a sequential-pay security as a line at the teller window: the top tranche is first in line for payments, then the second, and so on. All tranches might get scheduled payments of principal and interest, but prepayments (voluntary ones, like refis and property sales, as well as involuntary ones like recoveries from foreclosure or even put-back of a loan to the originator) are directed first to the topmost tranche until it is retired. That means that, all other things being equal, the first-to-be-retired tranche is least likely to ever lose principal. Of course, it means that the last-paid tranche becomes the bag-holder: by the time it gets eligible for principal payments there may be none left.

If you just left the credit risk issue as luck of the draw like that, you wouldn’t get far with the rating of the various tranches, nor would you get anybody sufficiently fired up about owning the bottom tranche. So this kind of CE is actually formalized in the security structure in a feature called “subordination,” or “senior/sub” structure. In essence, the tranches become liens on the underlying loans, and just like mortgages, each lien has a legal priority. Furthermore, the issue of loss timing cannot be just left to chance. Remember that, historically, most mortgage losses do not occur in the first two years of the loan. (Yeah, I know it’s different right now. That’s why what is happening right now is so scary.) Mortgages have a loss curve, and predicting that curve is hugely important and hugely difficult.

Besides the loss curve, you also have a prepayment curve—or vector—or explosion—or collapse—take your pick. Prepayments are as difficult to model as losses. Prepayments return principal to investors, so in that sense they reduce credit risk, but they don’t necessarily randomly represent the underlying pool. It is possible, indeed, it is likely, that the fastest prepayments in a given pool are the best-quality loans (those with refi opportunities). Insofar as there is significant variation of credit quality in the underlying pool, once we’re out of the GSE cookie-cutter business, you can end up with a pool that is paid down, but the remaining loans are the dregs. If you relied simply on prepayments to control credit quality, you’d end up with constant downgrades of the remaining tranches over time. That would not make these tranches attractive to investors, particularly to institutional investors who have to have, by law or bylaw, high-rated securities in their portfolios.

Therefore, what you generally get in a REMIC backed by first lien Alt-A or subprime is a combination of subordination, sequential pay, “overcollateralization” (OC), and excess interest. OC is simply a case of having an underlying pool that is larger than the face value of the security. If you issued a $100MM security backed by a $105MM pool of loans, you would have OC. The OC portion is sometimes called “equity,” and that’s what it is: just like in a mortgage loan, that’s the part that takes the first loss. Keep in mind that it’s not a separate pool. It’s not that there’s a special $5MM piece that does or does not experience loss. OC means that the issuer of the security did not get funded at “100% LTV,” as it were; it got funded (sold bonds to investors) for only 95% of the pool it created.

And like homeowner’s equity, OC is not constant: there is an initial OC amount, and also a “target” OC amount on these pools. OC can grow (and shrink). With REMICs, it doesn’t grow by having more loans added after the cut-off date (that can happen with some ABS home equity deals). It grows because that OC portion earns interest, and if that excess interest is not needed to cover losses, it can be used to increase OC.

Here’s a simple chart (I’m not an artist, you know) of an actual new Option ARM-backed REMIC issue I was looking at the other day. (I’m not here to encourage or discourage investing in anyone’s new issue, so whose issue this is isn’t the issue.) The credit enhancement percent on each tranche is the amount of lower-ranked principal that would have to be lost before the tranche in question took a loss; it’s the total of the lower-ranked tranches plus the OC divided by the pool balance.

The way this particular deal is set up, the monthly gross remittance (pool interest less servicing fee, master servicer fee, and lender-paid MI plus scheduled and unscheduled payments of principal) is applied first to the senior notes, on a pro-rata basis; then to the subordinate notes on a sequential basis (the highest-ranked subordinate note gets principal and interest payments until it is retired, then the next one down gets payments, and so on); and third to “excess cash.”

Each month, the “excess cash” is applied first to the OC, as principal, if the OC amount is under target; then to any realized losses of the senior notes; then to any realized losses of the subordinate notes, in order; then to the holder of the residual interest (the unrated interest of the security issuer). If in any period realized losses on the pool exceed the excess cash, then they are applied as a principal write-down first to the OC, then to the lowest subordinate tranche, then to the next-highest tranche, and so on.

In this deal, the “excess cash” is coming both from the interest on the OC portion and from excess spread on the collateralized portion—the rate paid on the underlying loans is greater than the coupon rate to the bondholders plus the fees. The actual payment structure on this deal is more complex than my summary would lead you to believe, and that is largely due to the fact that it’s backed by neg am loans. There are all kinds of “interest deferrals” and “carryovers” and a “final maturity reserve” and further weirdness that is necessary to deal with the underlying problem that the loans generate accrual but not cash-flow sometimes. I tend to wonder how many investors looking at this prospectus have any idea whatsoever about what it all means, but I have been called cynical. I do, however, have some experience with misunderstanding of the underlying loans, so there.

This deal also has a “step-down date,” which they all do. That is a date on which, if certain “trigger events” have not happened, the “excess cash” stops being applied to OC and can be released as distributions to the bondholders. In this deal, the triggers are 1) a serious delinquency rate in the pool of 40% or more of the total balance of subordinate bonds plus OC, or 2) realized losses of a certain percent or more as of a set of future dates. The step-down date calculation on my example pool is really complicated—they’re neg am loans—but generally it will be when the senior notes are completely paid down to zero principal. Remember that the senior notes get principal prepayments first, before the subordinate notes do, so they will mature faster. How fast will that be? Good question. Anybody issuing a new Option ARM Alt-A pool (this one is more than 75% stated income and most loans have a balance cap of 115%) in 2007 had better have a “Plan B” for dealing with very slow prepays. Don’t ask me what Plan B is.

So you can see how a tranche of this deal can get a credit rating that is much better or much worse than the rating of the underlying loans. You can also see how a downgrade of a tranche can happen. The credit rating of each tranche is not just a matter of the percent of credit enhancement under it, but it is strongly dependent on that, and if, say, sudden early losses on this pool wiped out the OC and ate into one or two of those subordinate tranches in a big way, you would see the senior notes and the top tier of subs downgraded, even though they may not have realized losses. They get the downgrade because they have less “support” than they were intended to have.

This one being a neg-am-backed deal, it wouldn’t surprise me any to see some subsequent downgrade if the loans negatively amortized a lot faster than initially predicted. That could create some evil cash-flow problems at the same time that any thinking person would conclude that the underlying loans were at greater risk of default, even if they hadn’t managed to default in abnormal numbers yet.

That’s important to keep in mind, because it implies that forcing a troubled loan to refinance (making it a prepayment, possibly with a prepayment penalty to goose the interest payment) will improve the cash-flow of this security (deferred interest will be replaced with cash interest), pay down the senior note balances (which will bring that step-down date, when the others finally get paid, closer), or supply some “excess cash” that can be used to cover current period losses. If you lowered the accrual rate on too many of these loans, you wouldn’t have that nice juicy “excess spread” to cover losses with.

On the other hand, if too many performing or even just under-performing but not yet loss-generating loans pay off too soon, you’re more likely to hit that “trigger,” because the percentage of seriously delinquent loans you’re left with gets up there. And, of course, if you refuse to modify but the borrower can’t get a refi from some other lender, you’re going to realize some losses. “You” are probably not a senior note holder in that case, but you could see your nice AAA drop down to AA if your CE gets whittled away too far. At the same time all this is going on, you have a servicer whose monthly servicing fee goes away when loans prepay, but who also has to keep advancing interest on delinquent loans until they are declared “uncollectable.” Your security can get a rating downgrade because your servicer’s got a cash-flow or capitalization problem, too.

The point here is that it is dangerous in the extreme to think that the interests involved here are either simple or uni-directional. We’ve seen Lewis Ranieri sounding the alarm over a security servicer’s inability to modify loans if it wanted to. It is not at all clear to me that all bondholders would want to, or that all classes would want to in the same degree or at a period in time that is mutually convenient. Conundrums are nasty things, as a certain Fed Chairman might have said once.

For those of you keeping score at home, notice that we haven’t yet gotten to the question of how a tranche of a REMIC becomes part of an asset pool backing a CDO. We’ll get there eventually. At the moment, just imagine the lowest-rated tranches of REMICs backed by some pretty scary loans becoming collateral for subsequent debt-funded securitization instruments that have even weirder and more complex structures than any known REMIC does. But don’t be frightened at the thought that we are getting to dangerous levels of derivation and dispersion and complexity, or some braver soul will call you a simple-minded cave bear and you’ll have to pretend that your feelings were hurt by that. I’ve got Kleenex if the tear-jerking gets out of hand, but I doubt the Calculated Riskers will need it.