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

Wednesday, May 23, 2007

Servicing Update: About Those Modifications . . .

by Tanta on 5/23/2007 07:00:00 AM

Fitch just convened a workshop of RMBS servicers, to find out what's on their minds, and the resulting Special Report, "U.S. RMBS Servicer Workshop," is available here. Anyone who wants further context for some of the questions being raised all over the place about servicing issues generally and subprime/Alt-A in particular should take a look; it's only five pages. My purpose this morning is mostly to continue a process I started in this post and fear I will be ending some time in 2009, but let's be optimistic. Per Fitch:

Servicers also discussed the factors contributing to increased defaults, including flat or decreasing home price appreciation; higher risk products where borrowers were qualified at teaser rates; affordability products; payment shock at ARM resets; high loan-to-value ratio (LTV) or piggyback loans; stated income; lower FICO scores and significant risk layering. Servicers indicated that as little as one year ago, refinances resulting in full payoffs were used most frequently for ARM resets and as a loss mitigation tool for defaulted loans. Today, refinance is not an option for many subprime borrowers due to tightened guidelines and flat or declining home appreciation. Further, third party and short sales, which resulted in some losses, although minimal due to some home price appreciation, are now less effective due to concerns on ultimate value in many markets, and while still less costly than real estate owned (REO) liquidation, are seeing dramatically higher losses.

Let's pause just for a moment to reflect on "refinances . . . as a loss mitigation tool for defaulted loans," or what mortgage wags have always described as "a rolling loan gathers no loss." Some loans were refinanced before they experienced delinquencies (borrowers looking for a way out just before that reset, or as the reset hit but while they could still manage to make the payment), and some were refinanced after they became delinquent. The latter requires a subprime lender to "take out" the loan by definition: it's currently defaulted. So this "cures" a problem from an MBS perspective only to the extent that it moves a problem into someone else's pool or portfolio (depending on what the refinancing lender chooses to do with the new loan).

Now, from the borrower's perspective, the refi may (or may not) actually improve things; that comes down to a question of whether the refinance terms represented a true benefit to the borrower or a long-term solution to the payment difficulties as opposed to, say, a temporary rolling into another short-term teaser to delay the inevitable while accreting yet another set of closing costs in the balance. You would never know which it is from any data provided on the original MBS delinquency report, because the loan is now "gone" from that pool. Hang onto that thought for a minute.
Almost all servicers indicated that they have not used modifications extensively as a loss mitigation tool in the past, primarily because there were other viable options for the borrowers. However, most indicated that they are preparing for significant increases in modification volume. . . . The servicers who had the most experience with loan modifications indicated that a re-default rate for modifications has been in the 30%–35% range, based on an admittedly small number of modifications completed to date. Discussions also included challenges around their modification strategies’ impact on transaction [i.e., the security pool] cash flows. Servicers generally expressed the belief that the ultimate loss to the transaction should be the only consideration in determining the execution of the best loss mitigation strategy. By using best execution calculations (lowest loss to transaction based on current value of property as compared to foreclosure and liquidation loss, which should be regularly updated using actual costs by area) applied on a consistent, documented basis, the actions of the servicer should be defendable to the consumer, their advocates and investors.

OK. I'm not yet prepared to take that 30-35% number as statistical gold, not just because of the small number of mods in question but because we don't know how representative the servicers in question or the loans in question are, how a "re-default" was defined, or how much the terms of the re-defaults were truly comparable to the terms of the successful loans. That said, in the current context it doesn't strike me as all that implausible.

The next question is what, for our present purposes at least, the real difference is between a mod and a refi as a "workout" tool. A very fruitful line of inquiry for someone with access to the data--O rating agencies, hear my cries!--would be to look at the "re-default rates" for those refis.

On the one hand, one might expect the failure rate for mods to be higher than for the refis, given that the "mod pool" involves some percentage of already-refinanced loans, meaning for at least some loans the mods are, actually, the "re-defaults" of the refis. On the other hand, consider that the "modifier" is someone who is acting out of an extreme sensitivity to loss calculations (there's nothing much on the gain side here) and who by definition gets to keep the "problem" if the mod doesn't solve it. The "modifier," in essence, is like our "traditional" hold-to-maturity lender. The "refinancer" is someone who is applying the same underwriting guidelines that got us into this mess in the first place, will make a profit on the transaction, and is undoubtedly planning to put the new loan into a new RMBS, which is why "refinancers" are in such short supply these days. In which hand am I holding the real uglies?

In the absence of any other data you have to guess that. It is highly likely that the final losses on a re-defaulted mod are substantially higher than they would have been had the servicer originally foreclosed. That does not settle the question of how much higher and whether the loss mitigation on the total mods is still in the plus column compared to what losses would have been if no loan had been modified. It also does not mean much at all unless good data on those failures is fed back into the "best execution" model that is used to make modification decisions in the first place. Servicers are going to say they've been doing that, duh. I'm cool with that; I suspect they have. That's why I keep scratching my head over the apparent absence of this data on the trustee reports. That actually came up in the Fitch workshop:
The discussion of investor reporting identified an increased need to develop standardized reporting for modified loans and detailed loan level default reporting. Some servicers indicated that in addition to the large variety of reporting standards and requirements, they are seeing increased requests for more granular reporting, indicating more heightened scrutiny on trustee statements.

Servicers expressed that they expect to see new deal documents which contain more detailed [information] reporting requirements, and most indicated that they would welcome this event. Some servicers indicated that they are prepared to provide this high level of detail, and had in fact attempted to provide it on current deals; however, because the documents did not require the additional fields, certain trustees were not willing or prepared to include this on their monthly reports to investors.

I find that last sentence absolutely believable, by the way. A mortgage servicing database is incredibly rich. No one I know has ever described a trustee report as "incredibly rich" unless the control group was some sticky notes. You would want to ask, at this point, why people buying bonds wouldn't create some demand for deal documents that require the trustee to make the kind of robust reports a servicer can easily produce. You would probably answer that the folks who are buying these bonds don't know enough about how a pig becomes a sausage to know what to ask for, and that, as usual, nobody gave a pork butt until the music stopped ("sausage" metaphors allow a high degree of mixing, so don't start with me). At some level it's just a matter of human nature to prefer the executive summary to the UberNerd reports, just as it is human nature to prefer "nobody told me" to "I failed to ask" when that strategy backfires.

This is why I'm shaking my head all the time over people who bought tranches of securities backed by a bunch of subprime refis (as well as those lovely 100% CLTV purchases made by the speculators' marks), as if that didn't represent a certain recycling of risk. Having done that, they're now a bit miffed to discover that they actually have to live with the "re-default" rate, since the refi door closed and now we have to modify. Honestly, how many times do you have to explain the rules of musical chairs to people? It's not a complicated game like Calvinball (although that might be what some market participants think they're playing).

In any case, I hope we can see the point our dear valiant mort_fin keeps trying to make in the comments, and that I keep failing to make clearly myself: a foreclosed subprime loan can represent the second, third, or nth subprime loan that borrower has had on the same property. In this case, the last loan in the series ended up in foreclosure; the first n loans ended up as prepayments due to refinances (or possibly mods, although that isn't reported as a new loan), which means they count in the "successful" pile of loans. Looking at current rates of delinquency or foreclosure cannot tell you anything about cumulative numbers, and it does not track one borrower all the way through all the refis until the end game. It is perfectly possible, at least hypothetically, to have a situation in which 40% of subprime homeowners eventually end up in FC or short sale or jingle mail, but only after three or four loans, so that on any given month, on the current total book of outstanding subprime loans, "only" 4% are currently in foreclosure.

Such a hypothesis could be tested with empirical data, but not with a mortgage servicer's current delinquency reports. If you wanted to use mortgage data, you would need something like OFHEO's house price index: a large database that has identifiers unique to the borrower, not just the loan or the property, that can identify "repeat transactions" as a chain, can distinguish between a payoff resulting from voluntary sale of the property that recoups the borrower's investment versus a payoff from refinance or distressed sale, and then can connect ultimate disposition of the home over some time horizon to the starting point (the original purchase money loan). Anybody who waves around these "but x% of borrowers make it" claims is blowing smoke if that claim is based on how many loans out of a mix of purchases and refis are in default at the moment. I am sorry to be so bloody repetitive but apparently you can't just shoot this one; you have to cut off its head, drive a stake through its heart, and bury it at the crossroads under a garlic wreath.

One last item from Fitch:
Higher defaults, and anticipated defaults due to the large volume of upcoming adjustable rate mortgage (ARM) resets, require an increase in loss mitigation proficiency and staffing levels. Low home price appreciation and tightening credit standards on new originations have reduced refinancing opportunities for borrowers so these loans will stay in portfolio longer and likely become delinquent. Regulatory and legislative scrutiny may also hamper effective timeline management. In addition, decreasing origination volume will change the ratio between performing loans and non-performing loans, which will be staying in transactions longer and are more costly to service. [my bold]

Two sets of numbers will be getting a lot of scrutiny in the coming months: mortgage-related job losses and non-performing loan ratios. Neither set of numbers will be easy to interpret in isolation from each other or from new origination volumes.

And another note: I did not actually make up that hypothetical 40% above entirely out of thin air. From Carolina Katz Reid's longitudinal study on low-income homeowners, published in 2004 and based on data that was current just up to the beginning of the most recent subprime lending boom:
Of low-income households who became homeowners, only 64 percent remain homeowners after 2 years, compared with 88 percent of high-income homeowners. Over 5 years, only 47 percent remain homeowners, compared with 77 percent of high-income homeowners. The hazard rate of leaving homeownership also varies by income group. For low-income respondents, the risk of returning to the rental market is extremely high in the first three years, but then drops off. The hazard for middle-income respondents is also high in the first three years, although lower than that of low income respondents. For high-income respondents, the hazard rate is more stable over time, hovering at around 5 percent per year. . . .

The dramatic finding is that for low income minorities, low-income whites, and middle-income minorities, the financial returns to homeownership over even 10 or more years of owning a home are extremely small. Indeed, for low-income minority homeowners, the average value of their housing only increased from $50,000 to $65,000—roughly a 30 percent increase over a 10 year period. While this does represent an increase in house value, this rate of return is less than the “riskless” return on Treasury bills.

Thanks, mort_fin, for the reference.