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

Saturday, April 14, 2007

MBS for UberNerds I: GSE Pass-Throughs

by Tanta on 4/14/2007 06:44:00 PM

We’ve had some fascinating and fun conversations lately about bonds and bondholders and stuff, and I first want to thank everyone who has contributed so much. I also observe that we do occasionally seem to be talking over or under or around each other, at times, about what mortgage-backed bonds are and how they work. There are places on the net you can get lots of technical information on this subject, but I’m not sure they are all pitched high enough to satisfy UberNerds or low enough to work for beginners, and in any case they lack periodic outbursts of editorializing, which is just boring.

Here, then, is installment one on the subject of mortgage-backed securities. I think the ones most people are interested in these days are the complex structured ones, but you really can’t understand what those are until you understand how the simple single-class ones work. Also, there’s a huge difference between a GSE-issued security and a private-issue one. Again, I think you need to understand the GSEs before you can really wrap your mind around the private-label stuff. So that’s where we start. As usual, if you’d rather eat toenail clippings than read something this long, by all means jump down to the next post. There are also lots of places on the web you can get the bullet-point version of this, and become appropriately dangerous; just Google it.

The oldest and simplest mortgage-backed security is the single-class “pass-through.” These are the specialty of Ginnie Mae and the GSEs. (The GSEs can and do issue structured securities, but their bread and butter is the pass-through. Private issuers can offer pass-throughs, although they’re much more attracted to the fancy ones.) To create one, the issuer acquires a pool of mortgage loans that meet all of its requirements. The underlying mortgages are serviced by someone other than Ginnie Mae or the GSEs—either the originating lender or a servicer to whom the originator sold the servicing rights, with the consent of the issuer. The pool of loans becomes a security by a kind of “swap” transaction: the ownership interest in the pool of loans is converted into a security—a bond—which can be sold on the secondary market.

The security has a “coupon,” which is the interest rate that the investor in the security receives. The “pass-through” concept is just that simple. Interest received on the underlying loans is passed through, on a pro-rata basis, to the investor, net of two or three things:

1. The servicing fee is retained by the servicer
2. The guarantee fee is retained by the issuer
3. Any lender-paid mortgage insurance or pool insurance is remitted to the insurer

The net result is the “coupon” of the security. For instance, one might have a pool of mortgages that back a security with a coupon of 5.50%. The underlying loans must, in aggregate, yield more than 5.50%. For example, assume a pool that has all low-LTV or primary MI policy loans, so we don’t have to mess with mortgage insurance at the pool level. They’re fixed rate loans, so the servicing fee is .25%. We’ll assume that they are of such high quality that the guarantee fee is 0.125% (it can be much higher than that). Therefore, the weighted average gross coupon of the underlying loans would have to be at least 5.875%, to make the bond coupon of 5.50%.

This doesn’t mean that every loan in the pool has a note rate of 5.875%, although it’s likely that 5.875% is the statistical “mode.” The underlying loans will have note rates in a “band” around 5.875%, so that the weighted average works out. (The average is weighted by loan amount.) Most simple pass-throughs aim for a fairly narrow band, because of the prepayment problem. High-rate loans pay off (refinance) very quickly when prevailing mortgage rates decline, and low-rate loans just sit there looking unprofitable when market rates rise. Because residential home mortgages always give the borrower the right to prepay, and because the GSEs don’t allow prepayment penalties in most cases (and if they do, it is only when the borrower is offered a lower note rate than market), the coupon of the security could fall pretty quickly if the underlying WA was based on half a pool of very high-yield loans and the other half of very low-yield loans. Achieving a band around the gross WAC of, say, no more than 1.00% plus or minus (200 bps total) keeps the prepayment characteristics of the underlying loans more uniform.

You can see, I hope, that there is already an “anti-predation” control here on a GSE pass-through (in terms of note rate, at least, if not fees and points). The GSE “bids” for loans to fill its desired current (par) coupon, the originator has to allocate loans to the pools within the required note-rate band, and there is thus a natural limitation on the originator’s incentive to produce loans that yield a great deal more than the market rate. (There can be excess yield here—it’s called “excess servicing” on the assumption that it is retained by the seller/servicer of the loans—but it does have those prepayment issues.) Of course, the GSEs can and do enforce other mechanisms for keeping predatory loans out of their pools, like refusing to allow prepayment penalties unless the borrower gets a rate cut in return. I’m simply making the point that certain forms of predation were less attractive to originators back in the day when they were selling all their loans to the GSEs in pass-throughs, long before the enactment of “high cost loan” laws. This in turn is one reason why the anti-predation laws are only really “catching up.”

So what does the guarantee on a pass-through mean? First, it means something different for Ginnie Maes than it does for GSEs. Ginnie Mae is an actual agency of the government, and its guarantee is therefore backed by the full faith and credit of the United States of America. That means that if a Ginnie Mae security suffered net losses (after FHA insurance or VA guaranty on the underlying loans) in excess of its guarantee fee collected, the payment of principal (and accrued but unpaid interest) would have to be covered by the taxpayer.

Fannie Mae and Freddie Mac are government-sponsored enterprises, meaning that they are private corporations (owned by shareholders) with a special government charter that gives them some advantages (tax relief, an emergency line of credit with the government in the event of disaster) in exchange for a government mandate and oversight. Their securities are not, therefore, backed by the full faith and credit: they are the sole obligation of the issuer. The bond market has always considered GSE securities to be “implicitly” government-backed, so GSE MBS trade a lot like Ginnie Maes, although there is some actual discount for the lack of the explicit full faith and credit backing.

There is also a difference in what, actually, is guaranteed here. First of all, Ginnie Mae does not actually buy the pool of underlying loans backing its securities. It “wraps” them with its guarantee, but does not invest equity in them; the loans are still owned by the originator/issuer. The GSEs can, although they do not have to, actually purchase loans outright for their pools. In any case, all Ginnies and Fannies, and all new Freddies, guarantee timely payment of principal and interest to the investor, even if it has not been collected from the borrower. (There are some old Freddies that guarantee timely payment of interest and eventual payment of principal.) So what does that mean?

“Timely” means as scheduled. If the underlying loans are, say, amortizing fixed-rate loans, then each is scheduled to pay interest and principal each month. Therefore, the guarantor makes sure that the investor receives its pass-through payments each month, even if the borrower did not make the payment. Partial or total prepayments of principal are “unscheduled,” and are passed through to the investor if and as they are made.

As a general rule, the GSEs require the mortgage servicer to advance payments on delinquent loans, so that the timely remittance to the investor can happen, and then to collect on those loans to pay itself off. Only when the loan gets to 90 days delinquent, usually, does the loan become “non-accrual” and the servicer’s obligation to advance end. The borrower still owes interest each month, but at this point the GSE (or the servicer, if such option is in the servicing contract) buys the loan out of the pool. This “prepays” the principal to the investor, which is now made whole, although it is no longer earning interest on that principal. If the GSE is now the owner of a delinquent whole loan, it directs the servicer to foreclose in the usual fashion, and is made whole out of the proceeds in the usual fashion, as is the servicer who advanced those delinquent payments. If mortgage insurance is involved, the MI pays its claim here, too.

So you can see that the GSEs are not, actually, guaranteeing that any investor is going to end up making any actual real money off the deal. The guarantee is that you will get interest payments as long as there is still principal invested to earn interest (that is, while there is still an underlying balance of mortgage loans), you will get it on time, and you will not lose your principal. If most of the pool prepays on you very quickly, via refinances or home sales or foreclosure (which, remember, is a prepayment to the investor), then you’ll need to reinvest that money elsewhere. If you paid a premium (a price above par) to buy the MBS in the first place, your actual yield could be zero or even less, but the GSEs do not guarantee that the MBS will be priced properly by you or anyone else. What the guarantee protects the investor from is credit risk, not prepayment risk, price risk, market risk, duration risk, etc.

It is important to understand that the GSEs do not hedge their credit risk—the credit risk they take off the investor—solely with a guarantee fee. In fact, the g-fee, as it is called, is for practical purposes the GSE’s “add on” hedge against its counterparty risk as much as the credit risk it thinks the underlying loans might have. A GSE seller/servicer with excellent performance and high net worth gets a better (lower) g-fee than its lesser-performing competitors. The g-fee is also driven heavily by the extent to which the loans underlying the pool are originated under standard GSE guidelines or any contract variances (typically referred to as “waivers,” more formally as “limited waivers of representations and warranties”) an individual seller/servicer might have negotiated with the GSE. For example, the standard rule for a given loan type and borrower might be that the CLTV cannot exceed 90%. A seller/servicer might get a waiver to originate some total dollar amount of loans with a CLTV of 95% with some other conditions (such as, for instance, a higher FICO than the required minimum in the standard program). This still requires the seller/servicer to rep and warrant that the waiver terms were not exceeded. In any case, the g-fee “prices in” such things.

But nobody, least of all the GSEs, pretends or claims that the g-fee is the total price of the credit risk or the timely payment guarantee cost. This is a very important and incredibly widely-misunderstood part of the whole thing, and it gets to the very heart of the issue with “nontraditional mortgages.” As I have indicated before, the GSEs control credit risk by making the pools diversified (in terms of geography, borrower type, etc.) and homogeneous in terms of the product type, the credit quality, the underwriting rules, the loan documentation, the legal documentation (the wording in the notes and mortgages), and a host of other factors including servicing rules. The combined “Selling” and “Servicing” guides published by Fannie and Freddie run to hundreds and hundreds of pages. There are rules for every damned little thing, as well as all the usual big things.

The idea is that what you end up with is a big honkin’ pool of loans with some variation in credit quality, but which results in acceptable average credit quality. After all, the whole point of owning an interest in a pool of loans, rather than owning a bunch of whole loans, is that they are “granular” rather than “lumpy,” and the investor is buying a pro-rata share of a lot of loans, not the total share of a few loans, which is what whole-loan investing is. I have used the old mortgage-bond-market metaphor of “sausage-making” before; this is what I’m talking about. The credit quality of a pool of loans is not equivalent to the weakest loan in the pool. It is a matter of the average of the loans in the pool, as long as the distribution or variation in credit quality is under control.

We talk a lot about representations and warranties, so let’s be clear on this. GSE MBS deals do not work by having originators of loans make up their own rules. What you represent and warrant is that you followed all the GSE rules on the loans (or any waivers you might have gotten). The GSE rules rule. Therefore, if subsequent to purchase, the GSE finds out you didn’t follow the rules—you made an untrue representation—you can be forced to buy the loan back (that is the warranty). The reason the whole thing works on a rep and warrant basis is just because it’s too big, there are too many loans, and it’s too expensive for the GSE to do several hours worth of due diligence review on each and every individual loan before securitizing it. It’s a kind of “trust, but verify” with some practical limitations.

That doesn’t mean the GSEs don’t do due diligence. They audit their seller/servicers, they sample loans for Quality Control review post-purchase, and they have a lot of up-front filters in place (largely in the loan underwriting and delivery software) to catch “out-of-scope” stuff. Again, a portion of the g-fee is based on the results of the GSE’s audit of the seller/servicer. Audit acceptability is a question not just of whether you tend to make true representations, but also how good you are for those warranties, should you have to cough up the cash.

Propaganda from certain other market participants aside, you cannot just put any old loan in a GSE MBS. You can’t do that even if you correctly represent up front that you just put any old loan in the pool: you must represent that any loan in the pool meets the GSE’s guidelines, and there is a very, very long list of loans they won’t allow. Some people seem to think the GSEs are the “buyer of last resort” for mortgage loans. That’s simple propaganda, folks. We can and ought to have a good conversation about the credit quality of what they do buy, but if you think they’re feeding at the bottom of the pond, you are mistaken. They take some risk—significantly on their “affordable housing” products, which risks they are mandated by the government to take. But there is no mortgage lender who does not take risk. There are just degrees of risk, and degrees of risk management.

This gets us to two related issues: “transparency” and “perfection” of the market price for credit risk. First of all, if the GSEs don’t have the time and energy to wade through every loan in the pool, you the investor certainly don’t have it, and you also lack the expertise. You can, if you want, go read 500 pages of GSE seller/servicer guidelines if you want to be all serious UberNerdy about it, but that isn’t how the bond market works last I checked. Most investors just accept the guarantee. Those investors who have a concern about whether the guarantee is reasonable, or who are concerned about their prepayment risk (that being an issue of the underlying loan quality and servicer competence as much as a question of market rates, as not all loans have the same opportunity to refinance or likelihood of prepaying via default), might put some effort into understanding what the underlying loans are all about.

But what is “transparent” to the investor are the GSE guidelines, and the reputation of the seller/servicer. Sophisticated institutional investors can do their own due diligence on that, as well. But bear in mind that the whole point of securitization is to “de-link” the originator of the loans from the issuer of the security. From the investor’s perspective, the net worth or vigiliance or competence of the seller/servicer is the GSE/issuer’s problem, since the GSE makes the guarantee. Can the GSEs make some bad bets there? Did we just read in the paper that Fannie Mae had to cancel a servicing contract with New Century? Why yes, we did just read that. From a public-policy or taxpayer-interest point of view, there may be plenty of reasons for you and me to get concerned about the counterparty or servicer risks the GSEs take, but from the simple perspective of our famous bondholders, it’s the GSE’s problem. That’s what the guarantee means.

The other issue is that, in short and simple terms, the investor in a pass-through MBS gets paid less interest than another privately-issued security might pay, because the MBS is credit-risk free. The price of the credit risk, you need to remember, is not equal to the g-fee; it is also a matter of the type of loans in the pool. In a sense, you can think of the price of credit risk as an issue of opportunity cost or resource allocation. By investing in a GSE MBS, you bought the kind of loans that the GSE can reasonably guarantee. That means you didn’t buy the kind of loans that the GSEs do not think they can cost-effectively guarantee. Back when the GSEs and Ginne Mae were the only games in town, that meant certain loans just didn't get made very often.

I have heard a lot of people calling lately for the re-imposition of the old requirement that every mortgage loan be a 30-year fixed with a 20% down payment. At some level, this is an “old requirement” in the sense that walking to school barefoot in the snow—uphill both ways—was an old requirement for educational transportation: less a fact about the old days than an artifact of a certain nostalgia for them. Were this to happen, in any case, investors in bonds would be buying MBS backed only by such loans, since those would be the only loans out there. There would be little issue with average credit quality if the weakest loan had to be better than average (the “Lake Wobegon” MBS).

It would certainly be cost-effective for the guarantor, which could improve yields to investors somewhat, if you assume that increased demand for fewer bonds doesn’t drive the yields back down to where they were or lower, and the increased hedge costs to the investor—who is no longer asking the borrower to pay it via an ARM—won’t end up in the (fixed) rate to the borrower somewhere. It would limit the returns available to bond buyers who might want to take a little more risk in return for a little more reward, since there wouldn’t be another other (mortgage) risk to invest in. Insofar as people who would be required to cough up 20% down payments were trying to save those up in bond funds, it could produce a certain conundrum.

The other side of the problem, as far as I’m concerned, are these voices demanding increased “flexibility” and “market share” and “modernization” of FHA, in particular. Remember that the issue isn’t just the reasonableness of the guidelines, it’s the homogeneity of the loans in the pool. “Flexible” is so often code for allowing a lot of variation on the bottom end that it’s impossible, for me at least, to imagine these proposals making any sense without significant increases in the insurance premiums paid by the borrowers to FHA and the g-fee paid by the lenders to Ginnie Mae. Unless we want to see the taxpayers actually buying out Ginnie Mae MBS holders for the first time in history. I notice no one cheering on these “modernization” proposals talking about the lunch tab, but perhaps I am simply inattentive.

Until such a day arrives that all loans must be more or less conservative than average, then, there will be opportunities for riskier mortgage investments. That gets us to the whole question of private-label and structured, as opposed to simple pass-through, securities, which will be the next installment.