FHFA’s CRT Report

On May 17 the Federal Housing Finance Agency (FHFA) published a report titled “Performance of Fannie Mae’s and Freddie Mac’s Credit Risk Transfer,” focusing on “the performance of the two key single-family CRT vehicles—securities issuances (also called capital markets CRTs) and insurance/reinsurance transactions—which together account for about 90 percent of all CRT issuance.” The report was remarkable in many respects, both for what it said and for what it did not say.

FHFA reminds us in this report that Fannie and Freddie’s CRT programs had their origins in the agency’s strategic plans for the companies in conservatorship: “In February 2012, FHFA released a Strategic Plan for Enterprise Conservatorships that identified several steps that FHFA and the Enterprises would pursue to ‘[shift] mortgage credit risk from the Enterprises (and, thereby, taxpayers) to private investors’….Pursuant to FHFA guidelines, Freddie Mac brought to market the first securities issuance CRT in July 2012, which FHFA’s then Acting Director described as ‘a key step in the process of attracting private capital back to the U.S. housing finance market.’ This was followed by Fannie Mae’s first securities issuance CRT offering in October 2012. Freddie Mac and Fannie Mae issued their first insurance/reinsurance CRTs in November 2013 and December 2014, respectively.”  

FHFA set annual targets for Fannie and Freddie’s credit risk transfers starting shortly after the beginning of the programs, and as a consequence CRTs have been used heavily by both companies for the past eight years (although Fannie has done no CRT transactions since the first quarter of 2020). In the report FHFA says, “Between July 2013 and February 2021, about $126 billion of risk in force (RIF), or the Enterprises’ maximum credit risk coverage, had been placed through securities issuance and insurance/reinsurance CRTs.”

I consistently have been critical of Fannie and Freddie’s credit risk transfer programs, arguing that they cause the companies to pay far too much for insurance against losses that have only a remote chance of occurring, and that investors will not buy new CRT issues during periods of financial stress, when they are most likely to be of value to the issuer. FHFA addresses both of these points in the May 17 report. On the second it says, “Concerns have been raised that CRT markets may be easily disrupted during periods of market stress, requiring the Enterprises to retain credit risk they had planned to transfer. The experience during and after the COVID-19 stress offers some support to these concerns.” The pandemic, of course, merely raised the threat of adverse credit performance—home prices in fact have jumped sharply since March 2020—which FHFA recognizes by noting, “CRTs remain untested by a serious credit event.”

But it’s the data on and the projections of the economics of Fannie and Freddie’s CRTs that are—or should be—the headlines from the FHFA report. First comes a straightforward accounting of the costs and benefits of the companies’ CRT programs to date: “As of February 2021, the Enterprises had paid approximately $15.0 billion in interest and premiums to CRT investors and counterparties and the Enterprises had received approximately $0.05 billion via investor write-downs and counterparty reimbursements,” with the added clarification that “To date, the only CRT tranches or layers that have incurred write downs or counterparty reimbursements (i.e., benefits to the Enterprises) have been in a few transactions issued in 2015, 2017 and 2018 in which the Enterprises sold first-loss tranches or layers.” FHFA next reveals the results of two performance simulations it asked a consulting firm, Milliman, to run, using what it called a “Baseline scenario” and a “2007 Replay.” In the baseline scenario, Fannie and Freddie’s lifetime CRT costs were $33.60 billion and their “ultimate benefits” (credit loss reimbursements) were $1.06 billion, for a net CRT cost of $32.55 billion, while in the 2007 Replay, lifetime CRT costs were $30.72 billion, ultimate benefits were $10.10 billion, and the net cost was $20.63 billion.

I’ll repeat those numbers, and emphasize that they come from FHFA, which has been directing Fannie and Freddie to do great volumes of CRTs for the past eight years: (a) the companies’ CRT programs to date have cost them $15.0 billion and returned $50 million in benefits (all on first-loss tranches); (b) the expected lifetime result of the companies’ CRTs on $126 billion of risk in force is a cost of nearly $34 billion and credit loss reimbursements of $1 billion, and (c) even in a worst-case scenario, Fannie and Freddie will pay $31 billion and only get $10 billion back in benefits, for a net loss of $21 billion.

I have long been saying that Fannie and Freddie’s CRT programs are noneconomic, but these actual and projected figures in the FHFA report are staggeringly noneconomic. I kept waiting for the authors to admit this, but they never did. Instead, they simply say, “FHFA continues to assess the CRT programs, including their costs and benefits as well as the benefits and risks to the safety and soundness of the Enterprises, the Enterprises’ ability to perform their statutory mission, and the liquidity, efficiency, competitiveness and resiliency of the national housing finance markets.”

What of any significance about Fannie and Freddie’s CRT programs could possibly be left to assess? We’ve known the common-sense objection to programmatic CRT issuance for some time. CRTs are priced by investors who seek a comfortable risk-adjusted return on their investment; if they sense that the mortgage market is becoming a little more risky they increase the returns they require on their CRTs, and if they sense the market may become a lot more risky they can withdraw from it altogether (as they did during the initial months of the pandemic). For these reasons, it would take systematic and persistent investor mispricing, and prolonged inattention to the leading indicators of mortgage credit loss, for the issuers of CRTs to ever get more out of them in credit loss transfers than they make in interest payments. And now we have eight years of historical data, along with prospective simulations overseen by FHFA, that lay out the terrible economics of CRTs with blistering clarity (while emphasizing this by noting, “CRT investors and counterparties are projected to receive a simple return of about 26 percent on the original reference pool UPB in the baseline scenario and 16 percent in the 2007 Replay”). There are no remaining mysteries about CRTs to be solved, but FHFA for some reason wants to pretend that there are.

This encapsulates what’s so wrong with the Calabria-led Federal Housing Finance Agency: they tout themselves as a “world-class regulator,” yet they seem to be operating in a world divorced from reality. Calabria and others at FHFA refer to CRTs as a triumph of “bringing private capital into the mortgage market,” but their staff has put out a report showing the truth to be just the opposite: by requiring huge amounts of annual interest payments while providing minimal absorption of credit losses in return, CRTs siphon tremendous amounts of capital out of the mortgage market, and weaken the companies FHFA regulates. Then there is the glaring contradiction between Calabria’s intense publicly-stated concern about the gap between Fannie and Freddie’s current level of capital and the (greatly excessive) amount he would like them to hold—which has led him to feel that FHFA must restrict the amount of high-risk business it permits the companies to do—and his repeatedly pushing them to issue CRTs that burn up their retained earnings and retard their capital growth.

There is something missing here. FHFA’s primary task is to regulate two companies that today take only one type of risk in one country on one type of asset (credit risk on U.S. residential mortgages), and it is sitting on a mountain of data that permit the quantification of that risk. But it seems not to know some of the most basic facts about it.

Since Fannie was put into conservatorship, it has published a chart in its quarterly Credit Supplement that shows the cumulative default rates of its single-family conventional book of business by origination year. Even a quick perusal of these charts makes clear that Fannie’s credit performance over the past two decades falls into three distinct groups: the years prior to 2004, 2004 through 2008, and the years after 2008. The 2004-2008 books are the outliers. And the difference between Fannie’s 2004-2008 books and the business it did before and after that period becomes even sharper when you include data on average loss severities, available in Fannie’s Connecticut Avenue Securities Investor Presentation (the default rate times the loss severity rate equals the credit loss rate).

Between 2004 and 2008, Fannie had a weighted average credit loss rate of 3.75 percent, driven by an 8.8 percent average default rate and 42.5 percent average loss severity. In contrast, during the previous five years—1999 through 2003—Fannie’s weighted average lifetime credit loss rate was just 45 basis points, the result of a 1.6 percent average default rate and a 28.0 percent average loss severity. And the credit performance of Fannie’s books from 2009 through 2020 is shaping up to be even better than that of the 1999-2003 books. The average “ever-to-date” default rate on Fannie’s books of business from 2009 through last year is well under 1.0 percent, while the weighted average loss severity of the 2009-2013 books (the only ones with enough seasoning for the data to be meaningful) has been 27.5 percent—virtually the same as the average of the 1999-2003 books. Thus, Fannie’s post-2009 books currently are performing consistent with a lifetime credit loss rate far less than the 45 basis-point average of the 1999-2003 books.

These are the data, but to turn them into a credit risk management strategy—whether for a credit risk transfer program or a regulatory capital standard—one needs an analytical overlay; specifically, is there any reason to think that the sharp deterioration in credit performance between 2004 and 2008, not just for Fannie and Freddie but for all mortgage lenders, was an anomaly, and not reflective of the inherent risk of residential mortgages?

In fact, there is, and I described it in a post titled “Some Simple Facts”: “In the late 1990s, the Federal Reserve under Alan Greenspan declined to regulate risky lending practices in the newly-emerging subprime market, favoring market regulation instead. Neither the Fed nor Treasury changed their regulatory stances when many of these practices began spreading to the prime mortgage market in 2003, nor when private-label securities (PLS) became the dominant means of secondary market financing for all single-family mortgages in 2004. In the absence of any prudential regulation, near-unlimited access to mortgages for unqualified borrowers through PLS issuance fueled an unsustainable boom in home sales, construction and prices that continued until the fall of 2007, when the PLS market finally collapsed. With PLS financing suddenly gone, and other lenders pulling back in an attempt to protect themselves, housing sales and starts plummeted, and home prices fell by 25 percent peak-to-trough before they could stabilize.”

I noted in the same post, “We learned from our mistakes. Post-crisis, the Fed and Treasury (and even Greenspan) admitted their deregulatory posture during the previous decade was an error. Congress in 2010 passed the Dodd-Frank Act, requiring lenders to apply an ‘ability to repay’ rule to mortgage borrowers and, through its qualified mortgage standard, effectively prohibiting the riskiest mortgage products and loan features that proliferated during the PLS bubble.” It is significant that almost all analysts who have examined the credit losses from Fannie and Freddie’s 2004-2008 books of business, including FHFA, have concluded that about half of them stemmed from products and risk features no longer permitted by Dodd-Frank. This means that even in the extremely unlikely event of a repeat of the 25 percent decline in home prices experienced during the Great Financial Crisis, the probable lifetime credit loss rate on Fannie and Freddie’s worst five years of business would not be 3.75 percent, it would be less than 2.0 percent—the average loss rate actually recorded on the 2004-2008 book, reduced by half to account for the absence of the products and risk features the company financed before the crisis, but no longer does.

FHFA should know this. Based on the historical data it has, it should know that a typical book of Fannie’s (or Freddie’s) single-family credit guaranty business is likely to have a lifetime credit loss rate of between 30 and 50 basis points, and that even a repeat of the severe home price declines of the mid- to late 2000s shouldn’t push credit losses on any book of business much above 200 basis points. Given that, it makes no sense to have credit risk transfers that don’t even kick in until after cumulative credit losses exceed 50 basis points, and go on to cover loss rates in excess of 400 basis points. Yet that’s how Fannie and Freddie’s CRT programs have been structured from the beginning. It’s no wonder that, as FHFA’s May 17 CRT report confirms, they’ve been such a colossal waste of money.  

Except the FHFA report doesn’t actually come out and say that Fannie and Freddie’s CRT programs have been a colossal waste of money; it says they deserve more study. Of course, they don’t, and that’s the problem with FHFA under Calabria. The facts say one thing, he says another, and as Director he gets his way, whether it’s on CRTs, business risk limits, or capital. Changing this counterfactual approach to Fannie and Freddie’s regulation very likely will require the Biden administration to change its FHFA Director.