As a follow-up to my proposal for the Urban Institute’s Housing Finance Reform Incubator series, which I called “Fixing What Works,” I’ve been examining whether having a credit guarantor use risk-sharing securities behind a layer of first-loss equity capital—or sandwiched between layers of equity capital—might provide credit protection more cost-effectively than equity capital used exclusively.
During the course of this exercise I read the 213-page prospectus for Fannie Mae’s July 2016 Connecticut Avenue Securities (CAS) risk-sharing transaction, to better understand the details of how these deals work. Much to my surprise, I learned that what are called the “mezzanine tranches” of Fannie’s CAS deals—the ones designed to take credit losses above a pre-set percentage absorbed by Fannie, up to another set percentage at which Fannie begins taking losses again—are intentionally structured to avoid taking losses. And Fannie discloses that in its prospectuses.
I never would have guessed this. To help explain its significance, I thought I would begin by giving some basic information about how the CAS structure works.
A typical CAS transaction has three component parts, or tranches. The first is the subordinate tranche, which absorbs a specified initial percentage of credit losses from the pool of mortgages it is protecting until its entire principal is gone. Fannie typically retains this tranche. In all five CAS deals done through the end of 2014, the subordinate tranche was equal to 0.30 percent—or thirty basis points—of the initial principal balance of the pool. That is enough to cover, for example, 1.0 percent defaults with 30 percent loss severity, or 1.2 percent defaults with 25 percent severity. In the four CAS deals issued in 2015, the first loss tranche was somewhat larger, either 0.40 percent or 0.50 percent, with Fannie still retaining all of it. So far this year, however, the subordinate tranche in each of Fannie’s four CAS deals has been 1.0 percent—and Fannie has sold an average of 28 percent of these first-loss tranches to third-party investors, who are being paid an average floating rate of LIBOR plus 11.55 percent to share this risk with Fannie on a pari passu basis.
The pricing on the recent first-loss tranches is quite generous, but the real story is the structuring and attributes of the next two layers—the mezzanine tranches. We’ll use the most recent CAS transaction, priced on July 19, as our example. Its three tranches—one subordinate and two mezzanines—provide what the prospectus calls “initial credit support” (note the word “initial”) of 4.0 percent for a $42.2 billion pool of single-family mortgages. These mortgages all have loan-to-value (LTV) ratios between 60 and 80 percent, with an average LTV of 75.7 percent and an average credit score of 748. The three CAS securities issued against this pool were: (a) $120 million in first-loss “Class 1B Notes,” which share with Fannie in taking all credit losses up to 1.0 percent of the $42.2 billion initial balance of the pool (Fannie’s piece of the 1B tranche, which it retained, was $302 million); (b) $701 million in “Class 1M-2 Notes,” which do not begin taking losses until all of the principal of the Class 1B Notes is exhausted, then take losses up to 2.75 percent of the $42.2 billion initial balance, and (c) $501 million in “Class 1M-1 Notes,” which begin taking losses when the 1M-2 Notes are exhausted, and in theory could take them up to 4.0 percent of the initial pool balance.
One hundred and forty pages into the prospectus for this transaction, there is a section titled “Prepayment and Yield Considerations,” intended to help investors evaluate the risks and potential returns of the notes they are being offered. It includes tables showing losses (called “write-downs”) taken by each of the three CAS tranches under 64 combinations of eight different annual default and prepayment rates. Default rates in the scenarios range from zero percent to 1.0 percent per year, while prepayment rates range from zero to 35 percent per year—a comprehensive span of potential outcomes deemed reasonable by Fannie, its investment bankers and their counsel.
The loss table for the 1M-1 Notes is an eye-opener. For this tranche, which is supposed to absorb losses that exceed 2.75 percent of the pool up to a maximum of 4.0 percent, there is no scenario, out of 64, in which it is shown to take any losses. Lest there be any question, a separate yield table confirms this: buyers of the 1M-1 Notes are shown as receiving the exact same yield—estimated to be 1.95 percent per year—in all 64 scenarios. Fannie is offering investors a “risk-sharing” security that according to its prospectus is risk free.
The sensitivity analysis for the 1M-2 Notes—which absorb losses above 1.0 percent of the pool, up to 2.75 percent—is almost as favorable. That tranche takes losses in only 9 of the 64 scenarios. Four of those have zero prepayments (which is not realistic), while the other five involve cumulative default rates between 4.6 percent and 7.8 percent. The estimated yield on the 1M-2 Notes is 4.79 percent in 55 out of the 64 scenarios, and it is negative (i.e., a net benefit to Fannie) in only three extremely unlikely ones.
My first reaction on seeing these loss and yield tables was disbelief. The mezzanine tranches couldn’t be that good for investors (and that bad for Fannie). But then I reviewed the average life tables in the prospectus, and it immediately became clear what was happening: the 1M-1 and 1M-2 Notes do not remain outstanding long enough to take what should be their share of credit losses.
This is a conscious design feature of the CAS structure. Even with no prepayments in the mortgage pool, the 1M-1 Note amortizes to zero by the end of its life. If there are prepayments—and there will be—after about a year’s worth of initial protection all prepayments associated with the subordinate and two mezzanine tranches are allocated first to the 1M-1 Note, until it is paid off. Next, prepayments are allocated to the 1M-2 Note until it is paid off. Only then do prepayments begin to pay down the subordinate tranche (the 1B Note held by investors, and the much larger 1B piece retained by Fannie). This feature of paying off the risk-bearing CAS tranches in reverse order of their loss-absorbing priority virtually guarantees that the 1M-1 Note will not survive to take losses, and makes it likely that the 1M-2 tranche will be able to provide only a small (or very small) percentage of its “initial” credit support.
As stated above, the July CAS 1M-1 Note does not provide any credit support until after losses exceed 2.75 percent of its associated mortgage pool. Losses of that magnitude take quite some time to develop. Fannie’s 2005 book of business, which had an 8 percent 10-year default rate, took ten years to reach 2.75 percent in losses. And Fannie’s 2006 book of business—which was acquired at the height of the home price bubble, was full of interest-only ARMs and no-doc loans, and had a 10-year default rate of over 13 percent—did not reach a 2.75 percent loss rate until its fifth year. The average life tables in the prospectus reveal that the 1M-1 Note disappears in a little over four years if its mortgage pool prepays at just a 10 percent annual rate, and at a 15 percent pool prepayment rate the 1M-1 Note barely lasts three years. Fannie’s 2005 and 2006 books of business prepaid at annual rates of between 15 and 20 percent. As bad as those two books were, the July 1M-1 CAS “risk-sharing” Note would not have reduced their credit losses by a nickel.
What are we to make of credit risk-sharing securities that take little or no real credit risk? Before addressing this question, I’ll add one more fact: as Fannie’s CAS series has progressed (the July 2016 deal was the thirteenth since October 2013), its successive risk-sharing tranches have provided less, not more, credit protection.
In each of the first five CAS deals through the end of 2014, investors in the Class 1M-2 and 1M-1 Notes were exposed to losses after Fannie absorbed the initial 30 basis points of loss. As a consequence, the sensitivity analyses in the prospectuses for these transactions showed the 1M-2 Note taking losses in 23 scenarios (out of 56—these prospectuses did not show results for the eight scenarios with zero defaults), while the 1M-1 Note took losses in 3 (twice) or 4 scenarios (three times). Fannie’s first-loss risk percentage was raised in the 2015 deals—first to 40, then to 50 basis points—but that did not materially reduce the number of loss-taking scenarios for the 1M-2 Notes (although with Fannie taking the first 50 basis points of loss the 1M-1 Note became loss free). Only by raising Fannie’s first-loss exposure to 100 basis points this year were the architects of the CAS structure able to reduce projected credit losses on the 1M-2 and 1M-l Notes to their current (miniscule) levels.
I think I know what’s going on here. At Treasury’s urging, FHFA has set a goal for Fannie to “transfer credit risk on at least 90 percent of the unpaid principal balance of newly acquired single-family mortgages in loan categories targeted for risk transfer.” Fannie has chosen CAS transactions as the main way to meet that goal. The problem, however, is that there are few if any natural holders of mortgage credit risk in the capital markets. Unlike entities such as Fannie, Freddie or the mortgage insurers, the primary capital markets buyers of Fannie’s CAS Notes—leveraged investors such as hedge funds and commercial banks—cannot count on being able to diversify their mortgage credit risk over time, and most have no independent means of evaluating it. Given these disadvantages, they price each CAS-type transaction to a worst-case outcome to ensure themselves an acceptable return. Fannie’s investment bankers know this, and to continue to sell to this ill-suited and cautious buyer base the large volumes of CAS Notes that Treasury and FHFA insist Fannie issue—at prices that seem reasonable—they’ve ended up creating mezzanine tranches that appear to transfer large amounts of credit risk but in reality transfer almost none.
Wall Street has every incentive to sustain this charade: it has a committed issuer (Fannie), a happy group of 1M-2 Note buyers (who so far this year have gotten securities with little risk, paying an average floating rate of LIBOR plus 550 basis points), and a very happy group of 1M-1 Note buyers (who have gotten securities with no risk, at a floating rate of LIBOR plus 190 basis points). The mystery is why Treasury and FHFA keep pushing what its prospectuses make clear is a Potemkin program. In a previous post (“Risk Sharing or Not”), I speculated that Treasury’s motive was to use CAS deals to try to develop a market for the sorts of risk-sharing securities it hopes can replace Fannie and Freddie as the primary means of financing mortgage credit risk. But that only makes sense if the CAS Notes really do share risk. There is no “market development” benefit from selling risk-sharing notes that transfer no meaningful amount of credit risk to anyone. Is it possible that neither Treasury nor FHFA understand how the 1M-1 and 1M-2 tranches actually work (or rather, don’t work)?
I reviewed the July CAS prospectus for insight into how a dollar of securitized risk coverage could be equated with a dollar of guarantor equity capital. I did indeed gain some insight. Equity capital for a credit guarantor is completely predictable; if you have two dollars of equity, you unequivocally can cover two dollars of credit losses. In contrast, the loss coverage of CAS-type risk-sharing securities is highly uncertain. It depends on the interaction of multiple elements, each of which is unpredictable—the amount and timing of mortgage defaults, the amount of loss severity, and the amount and timing of mortgage prepayments. As we’ve just seen, the July CAS 1M-1 and 1M-2 Notes claim to give coverage for up to three percent of credit losses on the pool of loans they were issued to protect, but at best their actual loss absorption will be a tiny fraction of that. One would be justified in assigning these notes an “equity equivalent” value of zero.
The egregious lack of loss coverage by Fannie’s 2016 CAS risk-bearing tranches can be partially remedied by adding more prepayment protection, keeping the notes outstanding longer so they can absorb more losses. But the broader issue will remain. Credit guarantors must be able to absorb tens of billions of dollars of losses with absolute certainty, and CAS and similar securities do not and never can provide that certainty. This fact, coupled with the revelation from the CAS program that the capacity of capital markets investors to take real, rather than pretended, credit risk is severely limited, should put to rest the notion that securitized risk-sharing ever could be a viable alternative to well capitalized credit guarantors as a source of secondary market financing for the $10 trillion U.S. residential mortgage market.
Finally, then, what should be done with the existing Connecticut Avenue Securities program? Unless Fannie’s investment bankers can radically restructure its 1M-1 and 1M-2 Notes so that they are capable of absorbing some reasonable amount of credit losses, Fannie should stop issuing these securities immediately. They are a sham, and a complete waste of the company’s money.
[Note: The original version of this post stated that the most recent CAS 1M-1 and 1M-2 Notes were issued in August; they in fact were issued in July. This version corrects that error.]