The Federal Housing Finance Agency (FHFA) received 37 responses to its request for input on single-family credit risk transfers. I have not read all of them (and likely will not), but the ones I have read appear not to have made the critical connection between FHFA’s request and the way it, in conjunction with Treasury, has been managing Fannie and Freddie in conservatorship for the past four-plus years. For both the current phase of the conservatorships—in which the companies are not allowed to retain capital—and for the future, FHFA sees credit risk transfers not as a supplement to Fannie and Freddie’s risk-taking but as a potential replacement for it. In not recognizing that fact, commentators missed an important opportunity to use their experience and knowledge to warn FHFA of the great danger of continuing to lead Fannie and Freddie towards an end-state in which credit risk transfers become the primary means of supporting their credit guarantees. That is a business model that does not and will not work.
The context for understanding FHFA’s view of risk-sharing mechanisms begins with its February 21, 2012 “Strategic Plan for Enterprise Conservatorships” (subtitled “The Next Chapter in a Story that Needs an Ending”). This plan, produced during the tenure of Acting Director Ed DeMarco, describes “the next chapter of conservatorship [as] one that focuses in earnest on building a secondary market infrastructure that will live beyond the Enterprises themselves… [and] will also see a gradual reduction in the Enterprises’ dominant position in holding mortgage credit risk as private capital is encouraged back into that role.”
Three initiatives were begun pursuant to this strategic plan to carry it out: requiring Fannie and Freddie to build a common securitization platform “capable of becoming a market utility,” increasing the companies’ guaranty fees to “move their pricing structure closer to the level one might expect to see if mortgage credit risk was borne solely by private capital,” and “establish[ing] loss sharing arrangements” to move credit risk away from the companies to capital markets investors. The goal of these initiatives—which have been underway for the past four and a half years—was to make it easier for Congress to at some point replace Fannie and Freddie with an alternative mechanism of secondary mortgage market finance.
There is no mystery as to what the mechanism envisioned by FHFA (and Treasury) as the replacement for Fannie and Freddie looks like. Two plans put out this year describe it in great detail, with only modest variations on the same idea. The first is the “More Promising Road to GSE Reform,” authored by five individuals associated with the Urban Institute and released in March (and updated several times since then); the second is “Toward a New Secondary Mortgage Market,” published this September by former FHFA Acting Director DeMarco and Michael Bright, both now at the Milken Institute.
Each proposal would replace shareholder-owned Fannie and Freddie with entities that use risk-transference mechanisms as their primary means of absorbing credit losses. The Promising Road would “merge Fannie Mae and Freddie Mac into a government corporation that is required to transfer all non-catastrophic credit risk into the private market,” while the DeMarco-Bright plan would “[r]econstitute Fannie Mae and Freddie Mac as lender-owned mutuals [after passing them through receivership], and build on the credit risk transfer (CRT) initiative to create a private market for mortgage credit risk.” The government corporation envisioned in the Promising Road would use risk-transfer mechanisms to provide capital up to 3.5 percent of its outstanding credit guarantees, and have another 2.5 percent in secondary capital in the form of noncumulative junior preferred stock. Similarly, DeMarco-Bright would use risk transfers to provide the first “300-500” basis points of capital, and as a secondary source of capital require the “owners of the mutual…to put up somewhere around 2 percent of additional enterprise capital in the companies.” Promising Road would use Fannie and Freddie’s common securitization platform to issue government-guaranteed securities, while DeMarco-Bright would use Ginnie Mae’s securitization platform to issue securities with Ginnie’s guaranty.
The underlying premise of both the Promising Road and DeMarco-Bright proposals is that “dispersing mortgage credit risk throughout the capital markets” is a less risky and equally cost-effective means of financing mortgages compared with leaving credit risk concentrated at too-big-to-fail specialized financial institutions (such as Fannie and Freddie). That sounds good in theory. But four years of FHFA-mandated experimentation with credit risk-sharing mechanisms by Fannie and Freddie leaves no doubt about the challenge confronting the Promising Road and DeMarco-Bright plans: both propose to use front- and back-end risk sharing to replace equity capital on a scale dwarfing anything attempted before, and in the face of overwhelming evidence of the limits of the market to make that capital available.
Using actual numbers helps put the risk-sharing challenge in perspective. Today, Fannie and Freddie finance $4.5 trillion in single-family residential mortgages. The Promising Road authors believe they need 3.5 percent of first-loss capital to back the single-family loans being made currently. (I think this is unjustifiably high, but it’s their proposal, so I’ll use their figure). DeMarco and Bright propose 300 to 500 basis points of first-loss capital, and to simplify I’ll use 3.5 percent for them as well.
To fully back $4.5 trillion of mortgages with 3.5 percent capital would require $158 billion. Where would that amount of risk-sharing capital come from? There are three potential sources: mortgage insurers and reinsurers, lenders keeping the credit risk on the loans they sell (through what are called recourse arrangements), and securitized credit risk transfers, or CRTs.
The combined capitalization of the U.S. private mortgage insurance industry is about $9 billion. Even were the MI industry to triple in size—which is not likely—this would result in only $18 billion in new risk-sharing capital, leaving $140 billion still to be found. Lender recourse won’t provide much of that. It has been around for decades, and seldom has been used because bank capital rules don’t favor it. (Bank capital standards are purely credit-based, so recourse arrangements have the same required capital as mortgages held in portfolio, which are much more profitable.) Almost all of the risk-sharing capital required by the Promising Road and DeMarco-Bright proposals, therefore, will have to come from securitized credit risk transfers. And that is where these proposals run into the wall of reality.
I’ve written extensively about the flaws and limitations of Fannie’s Connecticut Avenue Securities (CAS) risk-transfer program—in two previous posts titled “Risk Sharing, or Not” and “Far Less Than Meets the Eye” as well as elsewhere. Freddie’s Structured Agency Credit Risk (STACR) notes have the same flaws and limitations, which can be boiled down to the fact that neither CAS nor STACRs transfer much actual credit risk to their purchasers. This means that a dollar of CAS or STACRs is not equivalent to a dollar of equity capital, and to pretend that it is (or to ignore the fact that it isn’t), is to set the stage for a credit risk implosion down the road.
A good way to illustrate the concept of “equity equivalence” for securitized CRTs is to start with a structure that would be a very close substitute for equity capital. It would have a long final maturity—say 15 years. Its principal would not be reduced by mortgage amortization or prepayments, only if and when credit losses occur. The security would pay a floating rate of LIBOR plus a spread (determined at the time of pricing) on whatever principal is outstanding each month, and over the course of the security’s 15-year life investors would lose anywhere between none and all of their principal, depending on what happens with credit losses in the pool the security is insuring.
So why don’t we see risk-transfer securities that look like this? Because no mutual fund or pension fund investor would ever buy them. Their range of potential returns would be far too uncertain and volatile, and determined by a risk these investors do not understand (but the seller of the security does). Investors also would not be able to hedge the risk of these securities, and could diversify it only by purchasing the securities in large numbers (when they don’t want to purchase any). Investment bankers know this, so in order for them to sell the volumes of Fannie and Freddie’s CAS and STACR securities that FHFA has mandated, they have had to remove most of their credit risk by structuring them so that the majority of the principal pays off before the credit losses they are supposed to be insuring can be charged against it.
This is not mere theory; over the past four years we have seen investors’ credit risk aversion play out with Fannie’s CAS issuances. In the first five CAS transactions (through the end of 2014), Fannie absorbed only the initial 30 basis points of credit losses before transferring subsequent losses to the holders of the first risk-sharing tranche (called the “M-2”), which took losses up to 1.75 percent of the initial balance of the pool it was insuring, after which the next risk-sharing tranche (the “M-1”) took losses up to 3.00 percent of the pool (when Fannie began taking them again.) But just 30 basis points of first-loss protection caused losses to be transferred to the holders of both the M-2 and M-1 tranches too quickly for their liking. To induce investors to keep buying these tranches, Fannie has had to agree to take the first 100 basis points of loss. Even in a stress scenario, it takes time for that many loans to go bad, and during this time the M-1 and M-2 CAS tranches can pay off (without absorbing losses). As I noted in “Far Less Than Meets the Eye,” the loss sensitivity analyses in the prospectuses for Fannie’s last several CAS deals revealed that in none of the 64 combined prepayment and total credit loss scenarios did the M-1 CAS tranche take any credit losses, and the M-2 tranche took losses in only 9 of the 64 (four of which had zero prepayments, which is not realistic). In terms of “equity equivalents,” Fannie’s current M-1 tranches have no equity value at all—since they absorb no credit losses—and the M-2 tranches have relatively little.
So, in the real world, here is where we are. The Promising Road and DeMarco-Bright proposals require at least $158 billion in equity-equivalent credit-risk capital to meet their safety and soundness standards. Deeper front-end mortgage insurance and back-end reinsurance transactions might provide $20 billion of that. Lender recourse will not provide much at all, because any bank willing to put up capital for a recourse transaction can get ten times the return on that same capital by funding the loan on-balance sheet with low-cost insured deposits. And even after four years of CAS and STACR issuance, Fannie and Freddie have been unable to devise a risk-sharing structure that transfers any meaningful amount of credit losses to capital markets investors. The $17.6 billion in face value of CAS M-2 and M-1 tranches issued to date are worth at most $3 billion in equity equivalents, and the structures employed have become progressively less effective at transferring credit risk as more CAS tranches have been issued. Simply put, there is no credible path that gets anywhere close to the $158 billion in equity equivalents—or even half of that amount—required to make the Promising Road or DeMarco-Bright plans work as a substitute for Fannie and Freddie. They are theoretical fantasies.
In correspondence with several of the authors of the Promising Road proposal, I have spelled out in much more detail than given in this post (mercifully sparing the reader) why I believe their proposal is not workable. So far I have received no substantive refutations of any of the obstacles I’ve identified or the objections I’ve raised. The most direct responses I’ve gotten have been statements from one of the authors saying “I agree that the face value of a capital market CRT transaction isn’t necessarily equivalent to a $ of entity-based capital,” and “I agree that equity equivalency must be established. There is already work on this.” But I was given no specifics on what that work is, or what evidence exists of it being done. I was told only, “The CRT process as it currently stands is not sufficient. However, the CRT process is quickly evolving and I am confident that it will be up to the task.”
This, to me, sounds like the triumph of hope over experience. And it looks past the obvious solution. There is a mountain of evidence that capital markets investors will not take direct credit risk in anything like the amounts required to make the Promising Road or DeMarco-Bright proposals work. Yet these same investors will put private capital, and lots of it, into companies that themselves take the credit risk, then diversify and manage it—including laying off some to risk-sharing partners when it’s economic to do so, but keeping it otherwise—and pay stable and predictable dividends to the providers of that capital.
Hard equity capital invested up front, in entities responsible for preserving and earning a targeted return on it, is the only sure foundation for the secondary mortgage market of the future. That’s what I’ve proposed with my utility model of Fannie and Freddie, with updated capital standards to ensure they can survive a 25 percent nationwide drop in home prices, and regulated returns to remove any incentive to take excessive credit risk. Ironically, critics of the utility model reject it not because it doesn’t work (it has in the past, and it will work even better in the future with the changes I’m proposing) but because it’s “not enough reform.” Yet that is a hollow objection. Mortgage losses at commercial banks following the financial crisis were three times those of Fannie and Freddie, and we reformed the too-big-to-fail banks not by attempting to replace them but by having them hold more capital and regulating them more closely. That’s precisely what I’m proposing for Fannie and Freddie in my utility model—better regulation and more capital—and unlike the risk-sharing capital required by the Promising Road or DeMarco-Bright proposals, the equity capital needed by the utility model actually is obtainable.
In their current management of Fannie and Freddie in conservatorship, FHFA and Treasury have set their sights on a future risk-sharing model that has no chance of succeeding. And there is real danger in that. It will be very easy for advocates of this model to convince an uninformed public that securities like Fannie and Freddie’s CAS and STACRs really do transfer credit risk, because the experts say they do. But if that happens, the credit risk won’t disappear; it will stay hidden until a downturn arrives, then boomerang back on the issuer of the “risk-transfer” securities, who will be less able to absorb the resulting credit losses because they will have paid billions of dollars in wasted premiums to investors in securities that only pretended to take risk. The most effective way to minimize the chance of this occurring is to hold FHFA accountable for calculating, explaining and disclosing the equity capital equivalency of every CAS and STACR issue Fannie and Freddie do going forward, to prevent us from fooling ourselves that we’re transferring more credit risk than we truly are. The 37 respondents to FHFA’s request for input on credit risk transfers should have made that demand; they didn’t, because they didn’t know they needed to.