Private Mortgage Securitizations Have a Math Problem

According to Treasury Secretary Jacob Lew and many other others, the moribund market for private-label residential mortgage-backed securities needs reviving. Well, nobody should not get his hopes up, for one simple reason. The math doesn’t work. It hasn’t worked for 20 years.

More precisely, the original premise of PLS has not stood the test of time. The idea behind every PLS deal is that the arranger can study historical data and thereby predict how a single static mortgage pool will perform over its lifetime. His prediction must be accurate so that all investors, in a 100% debt-financed deal, get repaid. He has a one-time-only chance to get it right. There are no do overs after the deal closes.

Before getting into the weeds, it may be useful to recap a few other reasons why expectations of a major PLS revival should be lowered. First, most PLS sectors — subprime, alt-A, option ARMs — are gone forever. They proved to be unmitigated failures, and no long-term data suggests that they are viable, absent a housing boom. Only the jumbo sector, which represented less than one-third of all PLS issued from 1995 through 2007, has ever demonstrated viability.

Second, during the next few years, the volume of mortgage originations is going to be a lot smaller than it has been recently. The mortgage business goes through boom/bust refinancing cycles, which are driven by the Fed’s interest rate policy. We are currently in the bust phase. The next boom may be far off, since a homeowner with 30-year fixed-rate loan priced below 4.25% is unlikely to refinance soon.

Third, the common assumption — that any downsizing of the government sponsored enterprises translates into a commensurate rise in PLS demand — is a leap of faith, because PLS are profoundly different from GSE mortgage securities. The differences relate to structure, not the federal government. PLS deals involve the transfer interest rate risk and credit risk to the bondholders. GSE mortgage securities transfer interest rate risk but not credit risk; those deals all benefit from corporate guarantees, which preempt investor concerns about market timing or loan quality in any given pool. Plus, the numbers are irrefutable; GSE loan performance has always been exponentially superior to that of any other segment of the market, whereas PLS loan performance has always been the worst.

PLS pioneer Lewis Ranieri identified the math problem in 1994, when he gave a lecture at Northwestern Business School. “We have damaged the basic structure of the new housing finance system,” he said, referring to PLS in particular. “We did not build the system to finance refinancing. We built the system to finance housing.” Ranieri and others did not intend to build a system to finance refinancings; but the system mutated to do precisely that. Since 2001, the volume of refinancings has always matched or exceeded that of home purchase loans.

But Ranieri’s logic makes perfect sense, when you remember every static mortgage pool has a net present value, based on total projected net interest income and total projected credit losses. In the 1980s, when Ranieri helped build the PLS market, it was possible to study trends from the past to predict the future, because rate refinancings rarely occurred.

From the dismantling of Bretton Woods in 1971 until the early 1990s, 30-year mortgage rates never went down for any extended period. (The once-in-a-lifetime exception occurred in the early 1980s, when Paul Volker ended his anti-inflation campaign and slashed the Fed funds rate from its stratospheric highs.) Before Volker imposed his inflation medicine, 30-year rates hovered at around 9%. After Volker allowed Fed funds rates to plummet, mortgage rates hovered at around 10% until mid-1990.

Then Alan Greenspan’s Fed slashed rates for an extended period. In June 1990 the fed funds rate was 8.3%; in December 1992 it was 2.9%, the lowest it had been since 1963. Mortgage rates did not fall as dramatically, but they fell a lot. The 30-year rate fell from about 10% to 7%, and then hovered around 7.5% for quite a while. For the first time, refinancings exploded. Total refinancings were $70 billion in 1990; they tripled in 1991. From 1991 through 1993, refinancings exceeded $1.5 trillion. Eventually Greenspan started raising rates again, and refinancings dropped by 60% from 1993 to 1994. This was the first in a succession of convulsive boom/bust refinancing cycles that continued during and after his tenure at the Fed.

Remember that every static mortgage pool goes through a death spiral. With each successive month, the size of the pool shrinks, so that monthly interest income continually declines. Yet the average quality of loans remaining in the pool tends to worsen. Homeowners with good credit and positive equity prepay when it suits them; whereas homeowners without equity don’t prepay. The speed of the death spiral drives the NPV. More specifically, the speed of the death spiral determines if there will be sufficient net interest income to offset all credit losses from the pool.

These massive prepayment spurts can really screw up the NPV of a static mortgage pool, especially if they occur in the early years of a transaction. After the early 1990s, it was no longer possible to predict a static pool’s prepayment rate, because no mortgage model was clairvoyant as to Alan Greenspan’s future intentions.

Still, everybody, aside from PLS bondholders, wants to encourage refinancings. Balance sheet lenders and GSEs replace old loans with new loans, which bring additional upfront fees. Everyone in the originate-to-distribute chain wants more refinancings, because they all live off of upfront fees and commissions. Alan Greenspan wanted to stimulate the economy, especially when home prices were faltering in the early 1990s.

But if these refinancing booms screwed up a mortgage pool’s NPV, why did the PLS market keep rolling along for another twelve years, with no market disruptions?  There are several reasons why these issues could be swept under the rug.

Most important, the NPV volatility is concentrated in the most subordinate tranches. As you probably know, the most important thing for an investor in a PLS deal is seniority in the hierarchy of tranches. Just as the NPV of every mortgage pool is updated monthly, based on the latest loan performance data, the NPV of every tranche in every PLS deal is updated monthly. The NPV of a single mortgage pool may be volatile, but almost all that volatility is absorbed by the most subordinated tranches.

Going from the most senior tranche downward, each tranche relies on a decreasing level of overcollateralization, until you get to the very bottom tranche, the equity tranche, which benefits from no overcollateralization. The equity tranche (which is debt but collects any potential upside) relies entirely on excess spread — the difference between interest income collected from mortgages and the interest paid out to bondholders — to offset any credit risk. It also follows that, within a deal, the more subordinated the tranche, the more reliant it is on excess spread, and vice versa.

The subsequent refinancing boom, which began in 1997 and continued into 1998, did create big problems; it caused the subprime mortgage market to collapse. (For once, the boom was not triggered by Fed Policy but by fiscal policy; the Federal deficit came down so rapidly that long-term rates fell while the Fed funds rate remained stable.) In April 1997 the 30-year rate fixed rate was about 8%. By June of 1998 it was 7%, where it stayed until May 1999.

In the late 1990s, some long-forgotten subprime lenders — such as ContiFinancial, Southern Pacific — sold mortgages for PLS deals, in which they retained the equity tranches. In 1999, when faster-than-expected prepayments wiped out the NPVs of those deeply subordinated tranches, the companies’ equity was written down. Consequently, banks cut back on the credit lines used by subprime lenders to finance originations. And suddenly these subprime lenders were out of business.

Of course, these equity tranches were a small percentage of, what was then, a small niche in housing finance. And in the 1990s, the rated tranches of most subprime securitizations were guaranteed by mortgage insurance, so few investors took much of a hit.

In the late 1990s, credit losses were minimized because the economy was booming and home prices were rising everywhere. The California real estate boom, which began in 1997 and ended in 2006, skewed nationwide results. Loss severity on California subprime mortgage defaults was minimal; it fell from 5% in 2002 to 2% in 2003 and averaged 1% from 2004 through 2006. Contrast that with the subprime loss severity in Pennsylvania, which fell from 49% in 2003 to 26% in 2006. (In 2009 the average loss severity for California subprime loans was 70%. Timing is everything.) The timing of the California boom from 1997 until mid-2006 coincided with the roll out of the ratings agencies’ new mortgage models, which primarily relied on FICO scores to predict losses.

Though real estate is long-term investment, PLS can have economic lives that are very short. During the boom, most subprime PLS deals had average lives below three years. So if the excess spread earned during the few years is insufficient to cover all credit losses, the subordinated tranches will not recover principal.

We don’t know who purchased those subprime equity tranches after 2000, but we do know where most of the deeply subordinated tranches, which had investment grade credit ratings, ended up. They were stuffed into CDOs. Remember, any tranche rated below triple-A is deeply subordinated. Any non-triple-A tranche of subprime deal was subordinate to 80% of the capital structure; sub-triple-A tranche of an alt-A deal was subordinate to 90% of the capital structure; for jumbo deals it was 95%. Because these non-triple-A tranches were stuffed into CDOs at par, they rarely traded.

An NPV should foretell the future.  But, given the way that most PLS are structured, years can elapse before a tranche with a lousy NPV suffers a payment default. Most PLS are structured so that no principal is due and payable for 30 years. And mortgage prepayments generally assure that monthly cash flows can cover interest paid to bondholders.

So the ratings agencies simply allowed these zombie tranches, which would never recover principal but remained current on their interest, to languish for a long time before any downgrade was announced. The ratings agencies had no incentive to antagonize their customers, the Wall Street banks that hire them to rate structured deals. Also, the ratings agencies had no incentive to invite any criticisms of their methods. And so far as we know, none of the CDO managers expressed concerns about the bond ratings.

This basic problem — wherein excess spread can be wiped out quickly so that the values of the subordinated tranches are slashed overnight — has been swept under the rug for a long time. It’s hard to see how Treasury can enact reforms that overcome this problem.

David Fiderer has previously worked in energy banking for more than 20 years. He has written extensively about the financial crisis and is currently working on several projects concerning housing finance, including a book on the ratings agencies.