Mind the Gap

The “dog days” of August are a good time to put up a post on capital, credit pricing and cross-subsidization. It is not topical, so it can be read at people’s leisure. But it should be read, and carefully; its intent is to explain what needs to be done to restore Fannie and Freddie’s role in supporting affordable housing, to prepare for and facilitate their exit from conservatorship.   

For the past half-century, anyone taking the underground metro system (or “the Tube”) in London is likely to have heard a recorded message urging them to “mind the gap” as they disembarked at certain destinations. That gap is the space between the door of the train and the concrete edge of the platform, which varies by station.

Being in London last month caused me to recall that as CFO of Fannie I once was tempted to buy a passle of “Mind the Gap” t-shirts for the staffs of my Credit Finance group and the Single-family Business team, as mementos of our primary metric for balancing the market share, revenue, affordable housing goal and risk management objectives in our credit guaranty business—what we termed the “guaranty fee gap.” And as I thought about this, it occurred to me that a discussion of how the guaranty fee gap concept was used during my time at the company might be an understandable, “real world,” way to explain why the Calabria capital standard is so terribly damaging to the ability of Fannie and Freddie to finance any significant amount of affordable housing loans on reasonable terms, and why that standard must be replaced to restore their ability to do so.

Today, one rarely reads an article about the futures of Fannie or Freddie without hearing about the need to reform their “flawed business model.” Those flaws are never defined, because they are fictitious (as proven by the inability of the companies’ competitors and critics to devise a better model during fourteen years of trying). Fannie’s structure as a “private company with a public mission” in fact was remarkably efficient, and one of the most consequential reasons was its use of the guaranty fee gap as a management tool.

Fannie’s private incentives were to meet its shareholders’ expectations for earnings and growth; its public obligations were to meet (or exceed) its regulatory affordable housing goals and to protect the taxpayer from loss. When I was at Fannie, taxpayer protection was a given; we knew we had a valuable charter, which Congress could amend or revoke if we lost money (or made too much money relative to the volume of affordable housing business we did; it never occurred to any of us that Treasury might nationalize the company while it fully met its capital requirements). Balancing Fannie’s other public responsibility—meeting its affordable housing goals—with shareholder requirements for profits and growth was where our “gap management” came in. It worked then in support of affordable housing the same way it could and should work today, so it’s worth spending a bit of time to review.

We’ll start with credit pricing. To set a guaranty fee for any pool of loans—whatever its risk characteristics—you first must specify two key objectives: the return on capital you wish to earn under the most likely combination of future home prices and interest rates (which will be the central path of your home price model), and the cumulative lifetime loss rate you wish to protect against (or the “stress loss level,” which will be at the extreme, or inside, of the home price and interest rate paths generated by that pricing model). Once you’ve set those objectives, there will be only one combination of initial capital and guaranty fee that simultaneously earns your desired return on capital in the expected scenario, and (through a combination of initial capital and guaranty fees from loans that do not prepay) exactly covers your worst-case losses in the stress scenario you’ve specified. That will be the initial capital you should put up against this pool of loans, and the fee you should try to charge for guaranteeing them—the target guaranty fee.  

We did this when I was at Fannie. For all pools of single-family loans, the Credit Finance group would use its pricing model to determine the fee required to earn the company’s target return on capital, while passing the “loss stress test”. The difference, in basis points, between the fee charged by Single-family and the target fee set by Credit Finance was the guaranty fee gap; a charged fee higher than the target fee produced a positive gap, and a charged fee lower than the target produced a negative gap. The Single-family Business team had wide latitude on the size of the guaranty fee gaps on individual transactions, but we managed the gap at the aggregate level across all transactions each month.

When Single-family set a charged fee on any loan pool, we kept track of whether this fee was “gap positive,” “gap negative,” or neutral relative to its target fee, and by how much. We priced our lower-risk business with the deliberate intent to generate as large a positive guaranty fee gap as we could—without losing that business to a competitor (usually Freddie or a portfolio lender)—so that those excess fees (relative to our target) on a dollar basis could be applied to reduce the negative guaranty fee gap on our higher-risk loans, and thus enable us to do a much larger volume of affordable housing business on a profitable basis than we could have without this cross-subsidization. We set a limit as to how large an overall negative guaranty fee gap we would tolerate at the corporate level (in order to meet our shareholder objectives), then used our gap metrics to allocate the excess guaranty fees from gap positive business to higher-risk loans, in order to finance as many “goals-related” mortgages as we possibly could. It was an extremely efficient, and effective, process.

We all know that post-Financial Crisis, Fannie and Freddie will have to hold substantially more capital against their credit guarantees than before the crisis. But policymakers need to understand that the amount of capital required by the Calabria standard (formally the Enterprise Regulatory Capital Framework, or ERCF)—and equally importantly how this capital standard has been engineered to produce its “bank-like” risk-based percentage requirements—makes it impossible for the companies to use cross-subsidization to finance affordable housing in anything remotely comparable to what they used to be able to do, or could do again under a more reasonably constructed standard. It does so in three ways, none of which are related to risk: (a) ignoring that guaranty fees absorb credit losses; (b) arbitrarily increasing capital on all loans though add-ons, cushions and buffers, and (c) setting capital minimums on lower-risk loans, thus locking up capital that otherwise would be available to reduce the guaranty fees on higher-risk loans. These features combine to make the target fees on higher-risk loans unaffordable to most lower-income borrowers, and at the same time deprive Fannie and Freddie of the ability to use cross-subsidization to lower those fees by any meaningful amount.

We can gain insight into the limits of cross-subsidization from data Fannie and Freddie’s regulator, FHFA, published on the companies’ combined book of business in the spring of 2014, disclosing their required capital, target guaranty fees and charged fees by risk segment for the first quarter of that year. (This was at a time when former FHFA director Ed DeMarco had been raising Fannie and Freddie’s guaranty fees to “reduce [their] market share” and “encourage more private sector participation”; incoming director Mel Watt stopped those increases, but did not reverse them.) In the first quarter of 2014, FHFA was requiring the companies to price to average capital of a little over 300 basis points, and they set their target guaranty fees to achieve a return on capital of 11 percent after-tax (at what then was a 35 percent marginal tax rate). This made their average target guaranty fee for all business 72 basis points (including the 10 basis points they were required to charge and remit to Treasury, and still are), but they were able to charge “only” 60 basis points, leaving them with a negative guaranty fee gap of 12 basis points, and an expected after-tax return on capital of just 8.5 percent.

There are two aspects of these data that are instructive. The first is that there appears to be market resistance to average guaranty fees much above 60 basis points. (Fannie raised its average guaranty fee on new business from 57.4 basis points—again including the 10 basis points remitted to Treasury—in 2013 to 62.9 basis points in 2104, but its business growth stalled, and it dropped that average fee to 60.5 basis points in 2015.) The second is that with the 307 basis-point average capital required on Fannie and Freddie’s total new business in the first quarter of 2104, there was a large difference between the capital required on the lowest-risk quarter of that business, at 110 basis points, and the highest-risk quarter, at 570 basis points. (The middle fifty percent was capitalized at 275 basis points.) Because of these large capital differentials, there were comparably large differences in target guaranty fees, creating challenges for the companies’ cross-subsidization efforts.

Let’s assume that the 307 basis points of capital FHFA required of Fannie and Freddie’s first quarter 2014 books of business are the stress losses the companies must cover on today’s books (which are somewhat less risky), and that their target after-tax return on capital is not 11 percent, but 9 percent, at today’s marginal corporate tax rate of 21 percent. In these circumstances, the companies still could employ cross-subsidization reasonably effectively, if guaranty fees are counted as offsets to stress credit losses.

To cover 110 basis points of stress credit losses and earn a 9 percent after-tax return on the least-risky quarter of their business, they would need to put up 78 basis points of initial capital, and set a target guaranty fee of 30 basis points—composed of 10 basis points for Treasury, 8 basis points for administrative expenses, 4 basis points for expected credit losses, and another 8 basis points to absorb the 32 basis points of stress losses not covered by initial capital (using a multiple of 4 times the initial annual net guaranty fee—less than Fannie actually experienced on its December 2007 book that went through the financial crisis). For the middle 50 percent of their business, they would put up initial capital of 195 basis points and set a target guaranty fee of 42 basis points (to cover 80 basis points of stress losses), and for the riskiest quarter of business put up initial capital of 402 basis points and set a target fee of 64 basis points (to cover 168 basis points of losses). Even with target fees this high, charging an average of 38 basis points for the lowest-risk business would generate a positive guaranty fee gap of 8 basis points, and charging 44 basis points on the middle 50 percent would generate a positive gap of 2 basis points (on twice as many loans). This would enable the companies to charge as little as 52 basis points on their highest-risk loans (a negative gap of 12 basis points), making them much more affordable with no sacrifice to shareholder returns.

But note what happens when the three critical features of the Calabria standard get layered in. Not considering guaranty fees as offsets to credit losses—and requiring all of them to be covered by initial capital—increases the target guaranty fees on the lowest-, medium- and highest-risk business to 33 basis points, 51 basis points, and 81 basis points, respectively. And we still haven’t added the effects of the larger cushions and add-ons built into the risk-based capital standard, the stress and stability capital buffers, and the 20 percent minimum “risk-weight” (or 1.2 percent minimum capital requirement) on the lowest-risk loans. At June 30, 2022, these combined to raise Fannie and Freddie’s average required capital to more than 4 percent of total assets, a percentage that would have been higher were it not for credits for credit-risk transfer (CRT) securities issued, which cost the companies the equivalent of 4 basis points of foregone guaranty fees (3 for Fannie and 5 for Freddie).

We don’t have actual breakdowns of the companies’ June 30, 2022 capital requirements by risk segment, but we can approximate them. I estimate that the 1.2 percent minimum and the two buffers raise the required capital on the least-risky quarter of Fannie and Freddie’s business to around 225 basis points, which leads to a target guaranty fee of 50 basis points (including the 4 basis points for the CRTs issued to keep that capital amount from being even higher) for those loans. I also estimate required capital under the Calabria standard for the middle 50 percent of their business to be about 370 basis points—resulting in a target guaranty fee of 63 basis points—and the capital for the riskiest quarter to be about 660 basis points, leading to a target fee of 95 basis points.  

At these guaranty fee levels, cross-subsidization simply ceases to function. The most Fannie and Freddie were able to charge on the least-risky quarter of their business in early 2014, when their average total target fee was 72 basis points, was 50 basis points. Intuitively, it makes sense that there would be price resistance at this level. Charging more than that to insure the credit of loans with loan-to-value ratios of 65 percent or less—whose lifetime stress losses are less than one year’s worth of guaranty fee payments—risks this business being lost to portfolio lenders, who can earn 300 basis points of spread income on it while holding less capital than is required of the companies. But 50 basis points is now the target fee for Fannie and Freddie’s lowest-risk business under the Calabria standard. Without the ability to generate a positive gap on either this or their medium-risk business (which under the Calabria standard now has a 63 basis-point target fee), the companies have no way to reduce the average 95 basis-point fee on their riskiest loans, predominantly to affordable housing borrowers. What inevitably will happen, therefore, is that this riskier quarter will shrink to become far less than that, as borrowers can’t afford the full fee and the companies aren’t able to subsidize it.

Of course, all of this is an artificial problem. There is no economic need for Fannie and Freddie to have to hold anywhere close to 400 basis points of capital against their current books of business, with no consideration given to their charged guaranty fees. In its June 2018 capital proposal, FHFA said that the credit loss rate of Fannie’s 2007 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—through September 30, 2017 only would have been 150 basis points, which projects to lifetime credit losses of around 165 basis points. Even if Fannie were to stop operating tomorrow, its 45.9 basis-point average charged guaranty fee (net of the fee paid to Treasury) in the second quarter of this year would, after deducting administrative costs and the provision for loss, and at a four-times multiple, cover 136 basis points of those losses, leaving the need for just 29 basis points of initial capital. This is the reality of both companies’ current business, and it was confirmed by the results of the Dodd-Frank stress tests run on them by FHFA in 2020 and 2021.

Calabria’s capital requirement for Fannie and Freddie is not an economic standard; it’s a political one. Calabria was, and is, an ideological opponent of the companies. His capital standard was designed to do precisely what it does do: use gross overcapitalization to cripple Fannie and Freddie’s gap management and cross-subsidization processes, since he believes the government should not be intervening to make housing more affordable to lower-income borrowers. Unless the senior economic team of the Biden administration shares this view—which I doubt it does—it must scrap the Calabria capital standard entirely. Its structure has too many minimums, cushions and buffers that immobilize the excess capital on lower-risk loans, and also unduly penalizes higher-risk loans by not giving any value to their higher guaranty fees (which also prepay more slowly during periods of stress) as offsets to credit losses. For these and other reasons, the Calabria standard cannot be fixed.

And there is no need to try to fix it. In Capital Fact and Fiction, I say, “A rigorous and highly effective capital regime for Fannie and Freddie can be built with just three elements: (a) a true risk-based capital requirement based on a stress test run on each company’s book of business every quarter, with no cushions or add-ons; (b) a single ‘all purpose’ capital cushion, calculated as a percentage of this true risk-based requirement, and (c) a minimum capital percentage. Fannie and Freddie’s required capital would be the greater of the risk-based amount (plus the capital cushion) and the minimum percentage.” In the same piece I also point out that there is an obvious minimum capital percentage for the companies: 2.5 percent. This is the minimum requirement already in the Housing and Economic Recovery Act of 2008 for the companies’ on-balance sheet assets, and also is the minimum in the Calabria standard—without his (excessive and unjustified) “prescribed leverage buffer”.

Given the credit quality of Fannie and Freddie’s current books of business, a true-risk based capital standard would result in required stress capital of far less than 2.5 percent, making that percentage the companies’ binding capital requirement for the foreseeable future. At this amount of capital, they again would be able to “mind the gap,” and use their positive guaranty fee gaps from lower-risk business to significantly reduce charged fees on higher-risk loans, thereby allowing many more underserved borrowers to afford homes, without falling short of shareholders’ return objectives.

Also, a known 2.5 percent minimum capital requirement would open up, and facilitate, a path for the companies out of conservatorship. At June 30, 2022, 2.5 percent of Fannie and Freddie’s total assets (not “adjusted total assets,” a concept contrived by Calabria) was $185.9 billion. Were Treasury to cancel its senior preferred stock and eliminate its liquidation preference (as it should do), the companies’ core capital as of that date would be $90.4 billion. That is only a $95.5 billion shortfall to full capitalization—not too large to overcome with a combination of retained earnings and new equity issues in a relatively short time.

Fannie Mae and Freddie Mac have been in conservatorship for nearly fourteen years. They have been “conserved,” but remain constrained and captive by the misguided policies of previous administrations. Freeing them by revising their capital standards, canceling their (fully repaid) senior preferred stock, and returning them to private ownership will not be difficult for the Biden administration to do, and would be an unmistakable affirmation of its commitment to affordable housing. It must not allow that opportunity to go to waste.