Yesterday evening I submitted my response to FHFA’s May 15 “Fannie Mae and Freddie Mac Single-Family Mortgage Pricing Framework Request for Input” though the agency’s website.

This RFI discusses how the transition to the Enterprise Regulatory Capital Framework, or ERCF, this year has affected the companies’ guaranty fees (which are too high) and return on capital (which is too low). The RFI asks for public input on ten “topics,” but only the final one can fix the problems FHFA identifies: “Should risk-based pricing be calibrated to the ERCF?” I believe FHFA’s current leadership understands that the answer is “no,” but wants the industry to tell them this in their responses to the RFI, to provide political cover for lowering a capital requirement set by a predecessor Director, which nearly everyone thinks is excessive.

My RFI response makes the case for why and how the capital required of Fannie and Freddie should be reduced to no more than 2.5 percent of total assets. I recognize, however, that arguments from me won’t cause FHFA to change; the arguments that do will have to come from key stakeholders within the mortgage finance industry that FHFA knows and respects. To that end, I have submitted my response well before the August 14 deadline, to give other potential commenters ample opportunity to review it thoroughly, and draw from it whatever they may find useful in preparing their own submissions.

My response appears in full below.

On May 15, the Federal Housing Financing Agency (FHFA) issued a request for input (RFI) “soliciting comment on [Fannie Mae and Freddie Mac’s] single-family pricing framework and the goals and policy priorities that FHFA, as conservator and regulator of the Enterprises, should pursue in its oversight of the pricing framework.”

This RFI appears to be a response to criticism from the mortgage industry of the changes made by FHFA to Fannie and Freddie’s loan-level price adjustment (LLPA) grids, effective May 1, through which it lowered the LLPAs on certain higher-risk mortgages to improve their affordability, then offset the cost of those decreases by raising LLPAs on many lower-risk mortgages. In its RFI, FHFA defended these non-risk-based offsets by citing the impact of the transition in 2022 to the Enterprise Regulatory Capital Framework (ERCF) on the companies’ return on capital, noting: “Given the substantial amount of capital required to be held by the Enterprises on new mortgage acquisitions as a result of the ERCF, the Enterprises are not currently earning commercially reasonable aggregate returns on new single-family mortgage acquisitions. FHFA estimates that the Enterprises are generally earning mid-single digit returns on equity on aggregate new single-family mortgage acquisitions.”

While FHFA does not say this outright, it is evident that Fannie and Freddie are not now able to meet either of the two most critical imperatives of their charters: setting guaranty fees on their higher-risk loans at levels that are affordable to the borrowers who typically take them out, or earning “commercially reasonable aggregate returns” on their capital. To aid FHFA in resolving this dilemma, its RFI “seek(s) input on the process for setting the Enterprises’ single-family upfront guarantee fees, including whether it is appropriate to continue to link upfront guarantee fees to the ERCF, set risk-based upfront guarantee fees for both Enterprises, and set a minimum threshold for an Enterprise’s return on capital.”

Taking these in reverse order:

Minimum return on equity capital

The one parameter for Fannie and Freddie’s return on equity (ROE) over which FHFA has full control is the target ROE they build into their credit guaranty pricing models. And while I and other commenters may have opinions about what that percentage should be, the only opinions that matter are those of the investors who will be asked to provide the new capital required for the companies to ultimately exit conservatorship, or raise new equity during periods of stress if necessary. The financial advisers of FHFA, Fannie and Freddie are in the best positions to determine what this competitive market ROE is, and FHFA must seek, and heed, their advice.

FHFA also must distinguish between target and realized returns on equity. While the former can be controlled by FHFA, the latter are unknowable when guaranty fees are set, because they depend on the degree to which actual mortgage prepayments and credit losses—the key determinants of credit guaranty profitability—diverge from the medians for these variables produced within the pricing models. Once set, guaranty fees cannot be changed, making targets for realized returns on equity a futile exercise.

Finally, FHFA should not “hard-wire” different target (or maximum or minimum) ROEs for specific loan products or characteristics. Company managements should be permitted to decide on relative rates of return for different subsets of mortgage loans in real time and based on market conditions, subject to the constraint of neither significantly exceeding nor falling short of the target total ROE over a given time period (no shorter than one quarter). FHFA as regulator and supervisor should review and provide feedback on managements’ choices in this area—and sanction them for noncompliance if appropriate—but not micro-manage them.

Upfront versus ongoing guaranty fees

The mix between the upfront (or loan-level) and ongoing components of the guaranty fees Fannie and Freddie charge on their mortgage-backed securities (MBS) is not a significant contributor to the economic challenges the companies or their borrowers are facing.

Both Fannie and Freddie set target total fees for the MBS they guarantee. After the portion of the total fee to charge as an LLPA has been determined, those LLPAs are valued, in basis points, using the expected prepayment rates generated in the pricing models (and assumed in setting the ongoing fee). The only difference in the two components is that an LLPA has a known dollar value at the time it is charged, whereas the dollar amount of an ongoing fee depends on the realized prepayment rate—the faster an MBS repays, the lower the dollar amount of the ongoing fee, and vice-versa. This asymmetry in prepayment sensitivity has the potential to make a larger proportion of upfront fees on higher-risk loans an effective hedge, since greater credit losses typically (but not always) occur during times of falling home prices and lower interest rates, hence faster mortgage prepayments. Yet the impact on total credit guaranty profitability of overweighting LLPAs on higher-risk loans is not certain, and in any event decisions on LLPA weighting should be left to management.   

The answer to “should upfront guaranty fees be eliminated” is straightforward. If FHFA means eliminate them entirely, total guaranty fees would fall by an equivalent amount, and move the companies’ achievable ROE even further below a competitive market return. And if FHFA means “replace upfront fees with an equivalent amount of ongoing fees,” as noted above this would not make a notable difference either to the companies or to borrowers of affordable housing loans.

Link to Enterprise Regulatory Capital Framework

With Fannie and Freddie’s “commercially reasonable” ROE determined by the market, and no meaningful economic distinction between the upfront and ongoing components of their guaranty fees, the remaining avenue of inquiry in FHFA’s RFI that might provide it with a path for simultaneously reducing the companies’ guaranty fees on affordable housing loans and giving them a realistic chance of earning a market return on their capital is “whether it is appropriate to continue to link upfront [and total] guarantee fees to the ERCF.”

Here the answer clearly is “no.” FHFA’s risk-based capital standard for Fannie and Freddie never has been truly risk-based, and the ERCF literally is not risk-based at all; as detailed below, more than all of its 4 percent-plus required “risk-based” capital is the product of a layering of unrealistic assumptions, minimums, cushions, buffers and add-ons. FHFA at some point must replace the ERCF with a true risk-based capital standard, but in the short term it can significantly reduce the companies’ capital requirements, with no compromise to the goal of achieving an exceptionally high degree of safety and soundness, by removing the most arbitrary and damaging of the non-risk-based elements of the ERCF.

History of FHFA’s risk-based capital requirements

Over the past five years, FHFA has proposed three risk-based capital standards for Fannie and Freddie—the June 2018 standard, the May 2020 preliminary ERCF, and the December 2020 final ERCF. Each of them appeared, to one degree or another, to have been engineered to produce a desired percentage result, which over time became higher and higher, even as the results of the companies’ Dodd-Frank severely adverse stress tests were improving markedly. (To avoid noncomparability, the discussion below uses total assets as a common denominator for all capital percentages. FHFA’s May 2018 standard expressed Fannie and Freddie’s required capital as a percentage of total assets and off-balance sheet guarantees, which at September 30, 2017 were 1.04 times their combined total assets, while the final ERCF expressed capital in terms of a complex measure called “adjusted total assets,” which at June 30, 2020 was 1.09 times Fannie and Freddie’s combined total assets; the results of the Dodd-Frank stress tests are given as percentages of total assets.)  

The average risk-based capital requirement for Fannie and Freddie produced by the 2018 standard was 3.37 percent of the companies’ total assets at September 30, 2017. This was considerably higher than would have been predicted by the results of the 2017 Dodd-Frank tests, which was a loss of $35 billion, or 66 basis points of total assets, without a valuation reserve on deferred tax assets (and $100 billion, or 1.87 percent, with such a reserve). Yet a quick review of the structure of this standard revealed the main reason for the difference: unlike the FHFA and commercial bank Dodd-Frank stress tests (or episodes of severe stress in actuality), FHFA’s 2018 capital standard did not consider guaranty fees on loans that remained outstanding during the stress test as absorbing any credit losses, thus greatly overstating the losses to be covered by capital. FHFA called this feature “conservative.”   

In two years, the risk-based capital required of Fannie and Freddie by the 2018 standard had fallen considerably, to 2.42 percent of total assets as of September 30, 2019. Some of this decline was the result of improvements in the credit quality of new mortgages being guaranteed by Fannie and Freddie, but most resulted from non-risk-based anomalies in the standard. And by then, incoming FHFA Director Mark Calabria had not only announced his intention to promulgate a new capital standard for Fannie and Freddie, but also signaled, well before that standard was proposed, what he thought the companies’ required capital percentage ought to be, saying in an interview on Fox Business, “I think our objective over time is that you have capital levels at Fannie and Freddie that are comparable to other large financial institutions,” and that 4.5 percent capital was “kind of in the neighborhood of where we’re looking at.”

When the ERCF was first proposed, Calabria had added a 1.5 percent “prescribed leverage buffer amount” to the 2.5 percent minimum capital requirement of the 2018 standard, bringing total required minimum capital to 4.0 percent. He also changed the structure of the risk-based standard to bring Fannie and Freddie’s average required risk-based capital as of September 30, 2019 up to 4.13 percent of total assets, or 3.85 percent of the “adjusted total asset” measure he preferred. After receiving criticism that his 4.0 percent minimum capital requirement (which he also applied to “adjusted total assets”) should not be above the risk-based percentage, Calabria’s response in the final December 2020 standard was not to reduce the minimum, but to add more non-risk-based elements to the risk-based component, raising it to 4.65 percent of total assets (or 4.27 percent of adjusted total assets) at June 30, 2020—nearly double the 2.42 percent required of the companies by the June 2018 capital standard just nine months earlier. Neither Fannie nor Freddie needed any initial capital to survive FHFA’s Dodd-Frank severely adverse stress tests run in 2021.

A true risk-based standard

The concept of a true risk-based capital standard is simple. It is based on a stress test, in which a company’s book of credit guarantees is subjected to a simulated environment of falling home prices (and interest rates) of some specified amount. In Fannie and Freddie’s case, FHFA has chosen a replication of the stress environment of the Great Financial Crisis, in which average home prices nationwide fell by more than 25 percent peak to trough. The stress test is run on a liquidating book of business, which is conservative, because in an actual stress environment a company would at a minimum be replacing its runoff with new business, at equal or higher guaranty fees and with minimal losses in their initial years. The stress test reveals the amount of initial capital required to survive it, and to produce the full risk-based capital requirement an amount is added to account for model risk, operational and other risks, and provide a “margin of safety”—although not so much as to override the risk-based sensitivities of the standard. Typically, a minimum capital, or leverage, standard is then set so that the risk-based standard is binding on the company most of the time. 

With the ERCF, this capital setting process was reversed. Director Calabria began with a pre-determined minimum capital requirement of 4.0 percent, then engineered the risk-based requirement to produce a result that first was close to (in the preliminary proposal) and then above (in the final standard) that minimum percentage.

It is possible, however, to use historical and current data to approximate the amount of initial capital Fannie and Freddie would require to survive the stress they experienced during the Great Financial Crisis, were it to occur again. In its June 2018 capital proposal, FHFA said that the credit loss rate of Fannie’s 2007 single-family book of business “using current acquisition criteria”—that is, without the Alt A loans, interest-only ARMs and risk layering that resulted in over half of that book’s losses—would have been only 150 basis points through September 30, 2017. Using the pattern of Fannie’s actual credit losses, this would translate into a cumulative loss rate of about 115 basis points after the first five years of stress (at which point, during the crisis, Fannie’s revenues were again sufficient to absorb all current-year credit losses). In 2022, Fannie and Freddie’s average single-family net guaranty fee (total guaranty fee less TCCA fees paid to Treasury and 7.8 basis points of average administrative expenses) was 39.4 basis points. As going concerns, and without any business growth, an average net guaranty fee of 39.4 basis points would produce 197 basis points of cumulative net guaranty fee income in five years—far more than required to cover 115 basis points of cumulative stress losses over the same period, with no need to touch capital. The companies’ recent Dodd-Frank stress test results support this analysis.

It thus is not an exaggeration to say that more than all of the “risk-based” capital required of Fannie and Freddie by the ERCF at March 31, 2023—4.07 percent of total assets, or 3.70 percent of adjusted total assets—is a non-risk-based cushion of some sort, whether from highly conservative assumptions, imposed minimums, or add-ons. FHFA therefore has choices as to which of the non-risk-based aspects of the ERCF it can eliminate or pare back, in an amount sufficient to materially lower overall guaranty fees, permit efficient cross-subsidization between lower-risk and higher-risk loans, and allow Fannie and Freddie to earn market returns on the much lower, but still very conservative, amount of capital they will be required to hold. The four most significant elements of conservatism (or unrealism) in the ERCF, and their sources and degrees of conservatism, are addressed briefly below. 

Ignoring guaranty fee absorption of credit losses

Fannie and Freddie’s single-family stress tests are run on a liquidating book of business. Fannie has put out data that enables us to track the prepayment rates of its December 31, 2007 single-family mortgage book that experienced the stress of the Great Financial Crisis. (Freddie has not published comparable data.) Were Fannie and Freddie’s combined $6.51 trillion single-family books at March 31, 2023 to experience these same stress prepayment rates, their 39.4 basis-point average net guaranty fee would produce approximately $132 billion as those books liquidated over their lives. That equates to 175 basis points of the companies’ $7.54 trillion in combined total assets at March 31, 2023, and none of those fees are considered as absorbing credit losses in the ERCF version of the stress test (nor are any multifamily guaranty fees). Counting no guaranty fees as absorbing credit losses in a stress test is equivalent to assuming a 100 percent prepayment rate. That is not “conservative;” in a stress test, it is indefensible.

A conservative approach to the stress test would be to speed up the prepayment rate on Fannie’s 2007 single-family book from its observed average of 20.9 percent per year to, say, 25.0 percent per year. But even this still would produce $115 billion in net guaranty fees, or 153 basis points of the companies’ March 31, 2023 total assets. If the loss-absorbing value of these guaranty fees is ignored, it must, at a minimum, be acknowledged in assessing the merits or validity of any other elements of conservatism added to the risk-based standard.

Stress capital buffer

A similar point about ignored guaranty fees applies to the stress capital buffer, which the ERCF also refers to as a “going concern buffer.” This buffer is 75 basis points of adjusted total assets (79 basis points of total assets for Fannie and 86 basis points for Freddie at March 31, 2023), and its intent is “to ensure that the Enterprise would, in ordinary times, maintain regulatory capital that could be drawn down during a financial stress and still be regarded as a viable going concern after that stress.”

The concept of a going concern buffer is a valid one, but the ERCF misapplies it. If Fannie and Freddie are to be capitalized so as to remain going concerns—and they should be—then the ERCF needs to reflect the fact that their required capital has been determined on the basis of a liquidating book, without the replacement business a going concern will have. In an analysis I did of Fannie’s 2008-2012 stress experience, the company needed almost 100 basis points less initial capital to survive a going concern stress test compared with a liquidating book stress test, because the going concern stress period lasted only five years (until going-concern annual revenues exceeded annual losses), rather than thirty. And of course in a going-concern stress test guaranty fees must count as offsets to credit losses, since those fees will be there. Stating this point somewhat differently, because Fannie and Freddie’s risk-based capital requirement is based on a liquidating book of business, it already has a “stress capital buffer” built into it. There is no justification for a second one.

Capital minimums for lower-risk loans

The application in the ERCF of a minimum capital requirement of 1.6 percent to all loans, no matter how little credit risk they have, is not only unjustified (in light of all of the other conservative elements in the standard), it is particularly damaging to Fannie and Freddie’s abilities to assist low- and moderate-income homebuyers through cross-subsidization.

Without this minimum, the guaranty fees on the companies’ lowest-risk mortgages would have expected returns much higher than their overall target ROE, and those excess returns could be used to (invisibly) cross-subsidize higher-risk loans. But a capital minimum of 1.6 percent raises the required ROEs on lower-risk loans to the point where virtually all of the possibilities for effective cross-subsidization disappear. Cross-subsidization works very well when there is a close relationship between mortgage credit risk and required capital: the overall level of guaranty fees is reasonable, and a moderate degree of overpricing of lower-risk loans generates surplus fees that can be deployed to reduce the fees on certain higher-risk loans by significant amounts. But all of the non-risk-based cushions in the ERCF (including this minimum) cause guaranty fees on the lowest-risk loans to be set very high to begin with, which makes raising them further to cross-subsidize higher-risk loans much more likely to be perceived, and criticized, as happened with FHFA’s recent LLPA changes.

Stability capital buffer

The stability buffer is a non-risk-based, undisguised, capital penalty on Fannie and Freddie for doing more than a minor amount of business. It increases their capital by 5 basis points of adjusted total assets “for each percentage point of market share exceeding the threshold of 5 percent of total residential mortgage debt outstanding.” The stability buffer added 104 basis points of total assets (or 99 basis points of adjusted total assets) to Fannie’s capital requirement, and 71 basis points of total assets (66 basis points of adjusted total assets) to Freddie’s, as of March 31, 2023.

Fannie and Freddie are the only two companies in America who guarantee the credit of conventional residential mortgages. Their guaranty makes residential mortgages attractive to contractual investors like mutual funds, pension funds and insurance companies, who lack the ability to manage the credit risk of individual mortgages but are well suited to manage the interest-rate risk of 30-year fixed-rate mortgages in MBS because they do not use leverage; the sources of their funding are stable and predictable. The main alternative to a contractual investor managing the interest-rate risk of a Fannie or Freddie-guaranteed MBS is a commercial bank holding an unsecuritized mortgage in its portfolio. For at least the past three decades, FDIC data show that banks’ serious delinquency and default rates on residential mortgages average about triple those of Fannie and Freddie, and periodically many banks fail because of funding 30-year fixed-rate mortgages and other long-term assets with short-term consumer deposits and purchased funds—most recently, Silicon Valley Bank, Signature Bank and Republic Bank. It simply is incorrect to assert that higher “market shares” of Fannie and Freddie increase systemic risk, and must be discouraged by imposing a capital penalty linked to their size. The misnamed “stability capital buffer” in fact encourages the opposite outcome.

Recommendation

Fannie Mae and Freddie Mac’s guaranty fees—particularly on loans to affordable housing borrowers—are far too high, and their current and prospective returns on equity are far too low, for one blatantly obvious reason: their “risk-based” capital required by the ERCF is in no sense risk-based. At some point, FHFA must scrap the overly complex ERCF, and start over in a way that deals with the original weaknesses of the 2018 standard (including not counting guaranty fees as offsets to credit losses, using current rather than original loan-to-value ratios to determine risk-based capital, and treating credit-risk transfer securities too generously). But in the short term, the two critical problems FHFA identifies in this RFI of unaffordable guaranty fees on higher-risk loans and an insufficient return on Fannie and Freddie’s capital can be solved quickly and simply if FHFA changes the ERCF to:

• Eliminate the prescribed leverage buffer in Fannie and Freddie’s minimum capital requirement, reducing it to the 2.5 percent in FHFA’s 2018 capital proposal, but as a percentage of total assets, and

• In the risk-based standard, eliminate the stability capital buffer and the 1.6 percent capital minimum, and if necessary lower the stress capital buffer by enough to reduce Fannie and Freddie’s required risk-based capital to or below the 2.5 percent minimum, leaving more than enough of the ERCF’s remaining conservatism and cushions in place to ensure an unquestionably high level of safety and soundness for taxpayers.