The Right Choice on Capital

One of the recommendations of the “Blueprint for Restoring Safety and Soundness to the GSEs” released earlier this month by the investment firm Moelis & Company is the imposition of “rigorous new risk and leverage-based capital standards” on Fannie Mae and Freddie Mac. These include two fixed percentages. The first is a 4.25 percent minimum taken directly from the Basel III bank capital standards: 8.5 percent equity capital with a 50 percent risk weight for residential mortgages (which is risk-based in name only). The second is a “two-tiered leverage ratio,” made up of 3 percent equity capital and another 2 percent that can be “soft” capital such as credit risk transfer (CRT) securities, for total capital of 5.0 percent. Both versions would apply to all of Fannie and Freddie’s business, irrespective of its risk. Because the companies would add the interest cost of CRTs to their guaranty fees, pricing using the 5.0 percent leverage ratio and the 4.25 minimum should be about the same: Fannie and Freddie would likely charge close to a 70 basis point average guaranty fee (before the 10 basis point payroll tax fee paid to Treasury) in each case.

The rationale Moelis gives for this recommendation is that, “These illustrative capital requirements…are broadly consistent with approaches applied to other large financial institutions, and represent reasonable estimates of capital standards for the Enterprises.” The “other large financial institutions” are of course big banks, and in prescribing bank-like capital standards for Fannie and Freddie Moelis falls solidly in line with most other reform proposals. To its credit, however, Moelis also goes on to say, “Given the unique nature of Fannie Mae and Freddie Mac’s businesses, and particularly the scale of their mortgage guarantee businesses, FHFA may elect to implement a more nuanced risk-weighting system
 for mortgages, as compared to the fairly simplistic…approach applied to multi-product banks,” adding, “A more nuanced approach…could also help to broaden the ‘credit box’ that has historically excluded large groups of deserving Americans from obtaining a mortgage.”

Moelis’ instincts about the merits of a “more nuanced risk-weighting system” for Fannie and Freddie’s capital are correct, and I believe that once the firm does a more comprehensive analysis of the pros and cons of a true risk-based capital standard for the companies compared with a fixed ratio-based standard it will argue much more forcefully and convincingly for the former.

It is not possible to have a true risk-based capital standard for commercial banks, because their business spans numerous products and markets with multiple types of risk, many not statistically predictable. But such a standard is possible for single-product mortgage guarantors, who take residential mortgage credit risk for which vast amounts of historical performance data exist. To implement it, Treasury (or whoever the administration designates) first would pick the stress environment it wants the guarantors to be able to protect against. FHFA then would use Fannie and Freddie’s historical data (for loans they currently are allowed to acquire, that is, excluding the interest-only ARMs and “Alt-A” mortgages that caused half of their credit losses in 2008-2012) to project stress default rates and loss severities by risk category—at a minimum, combinations of loan-to-value (LTV) ratios and credit scores. The resulting loss amounts would be the basis for determining Fannie and Freddie’s new capital requirements, set at the risk-category level. I’ve recommended that FHFA also set a minimum capital requirement, derived from the individual risk-category capital requirements and a mix of business deemed prudent but also encompassing broad service to affordable housing borrowers.

The argument made against a true risk-based capital standard for Fannie and Freddie is that it would give the companies an unfair capital advantage over banks, which have to meet a ratio-based standard. Yet subjecting the companies’ credit guaranty business to the same fixed capital ratio as banks (whether 4.25 percent or some other figure) in the guise of a “level playing field” ignores the fact that the Basel III risk weights for banks do not differentiate between credit risk and interest rate risk, whereas Fannie and Freddie’s minimum (and risk-based) standards have in the past and almost certainly will in the future.

Commercial banks today hold $3.7 trillion of single-family whole loans and MBS on their balance sheets, $2.2 trillion of which have long-term fixed rates. They fund the vast majority of these mortgages with short-term consumer deposits and purchased funds (or “hot money”). Banks’ 4.25 percent Basel III capital requirement for single-family mortgages has to cover not just credit risk but also the risk of financing 30-year fixed-rate and capped adjustable-rate mortgages with short-term debt. In Fannie and Freddie’s regulatory standards, those risks are capitalized separately; there is both a 45 basis point minimum for guaranteed mortgages (now universally viewed as too little) and a 250 basis point minimum for mortgages held in portfolio. The companies have 205 basis points of required interest rate risk capital for on-balance sheet mortgages even though they can and do match their durations with a combination of long-term debt and derivatives, and “rebalance” to keep those durations matched as interest rates change.

Banks take much more interest rate risk in their mortgage portfolios than Fannie and Freddie do. To create an actual level playing field on capital between banks and Fannie and Freddie’s single-family credit guaranty business, therefore, at least 2.05 percent, and realistically considerably more, of banks’ 4.25 percent Basel III capital charge must be allocated to that risk, leaving at most 2.2 percent as the amount applicable to the credit risk in Fannie and Freddie’s guaranty business. Applying banks’ fixed capital ratio directly to the companies’ credit guarantees without this adjustment would result in Fannie and Freddie’s portfolios being capitalized at 6.30 percent (4.25 plus 2.05 for interest rate risk), which clearly is not level with banks.

If we take away the level playing field argument—as we should—no good reason is left for rejecting a true risk-based capital standard for Fannie and Freddie. A single ratio-based standard imposed on any guarantor’s entire book of business creates a capital incentive to take uncompensated credit risk and a capital disincentive to use cross-subsidization to more favorably price affordable housing loans. A true risk-based standard does the opposite. In addition, unlike a fixed capital ratio a true risk-based standard produces a clear framework for evaluating credit risk transfer securities to insure against unexpected loss, and for attracting and pricing private reinsurance to cover catastrophic loss.

Let’s begin by examining how a fixed-ratio capital standard distorts credit pricing decisions and thus a credit guarantor’s risk profile. Capital needs to be related to risk, and with a fixed capital ratio what happens is that required capital ends up determining the risk profile, rather than the risk profile determining the capital.

An example may help to explain this. In 2014, FHFA published a capital and pricing grid for all of the business done by Fannie and Freddie during the first quarter of that year, broken into nine combinations of LTV ratio and credit score ranges. There was a wide disparity between the lowest- and highest-risk 20 percent of that business. For the lowest-risk 20 percent, FHFA calculated economic capital of 115 basis points and target guaranty fees of 25 basis points, compared with economic capital of 600 basis points and target guaranty fees of 120 basis points for the highest-risk 20 percent. For all of the business combined, FHFA’s calculated average economic capital was a little over 300 basis points, and average target guaranty fees were a little over 60 basis points.

FHFA did not disclose how it came up with these economic capital figures or target guaranty fees, but for this exercise let’s assume that the capital numbers represent FHFA’s best estimates of capital required to pass a rigorous stress test, and that the guaranty fees are consistent with a reasonable return on that capital. Let’s further assume that all of Fannie and Freddie’s business is subject to a 300 basis point fixed capital requirement, irrespective of risk. Now we’ll ask: how would the companies set their guaranty fees for the loans in FHFA’s very different risk categories?

I can imagine facing this challenge, having done risk management for a credit guaranty business for two decades. With a fixed 300 basis points of capital we would need to charge an average guaranty fee of 60 basis points to earn our target rate of return. We wouldn’t be able to charge that on the lowest-risk 20 percent of our business (with a 25 basis point target fee) without losing most of it, but if we cut fees to get more our required capital won’t change and we’d fall short of our return target. So what do we do? Do we cut fees on our lowest-risk business from 60 basis points to, say, 40, then try to recoup that discount by charging not 120 basis points but 140 basis points for our highest-risk business? That would make credit guarantees even less affordable for the borrowers who take out these loans.

What I believe would very likely happen if regulators insist on putting a flat capital charge on a single-product business with variable categories of risk is that guaranty fees for all risk categories would stay relatively close to the average (in this example, 60 basis points), and the credit guarantor upon whom this capital regime has been imposed will end up with a significantly riskier book of business—with overpriced lower-risk business staying away and underpriced higher-risk business being attracted—than the fixed-capital ratio was intended to have produced. That’s the opposite of what a safety and soundness regulator should want.

With a true risk-based capital standard, in contrast, the credit risks of all types of business automatically get capitalized at the proper level, and the pricing dynamics encourage, not discourage, a balanced and diverse book of business, including loans for affordable housing. Guarantors would be able to employ a version of the cross-subsidization approach used effectively by Fannie and Freddie prior to the 2008 collapse: charging somewhat more than an economic guaranty fee for lower-risk business (although not so much as to lose a meaningful amount of it), and using the resulting overages to selectively lower fees to obtain higher-risk affordable housing business, while still meeting their return targets and passing their stress tests.

Another crucial benefit of a true risk-based capital standard is that it increases the efficiency and reduces the cost of third-party credit risk transactions by giving the concepts of expected loss, unexpected loss and catastrophic loss—which are either vague or arbitrary under a ratio-based standard—reliable numerical values. This enables proper economic analysis and market pricing of both credit risk transfers (CRTs) and catastrophic risk insurance.

Some “real world” detail is useful in explaining this point as well. When a credit guarantor guarantees the performance of a pool of mortgages, it does so without knowing the exact economic, interest rate or home price environments this pool will experience over its life. For that reason, the guarantor looks at a wide variety of possible outcomes for these variables, and prices to expected values.

When I was at Fannie we used 500 calibrated combinations of future interest rate and home price changes to project prepayments, defaults and loss severities for the mortgage pools we insured, based on their risk characteristics. Our expected loss was the average loss in all scenarios—usually in the 3 to 6 basis point range for our entire book, but considerably higher for riskier pools of loans—and it went into our pricing build-up, along with our administrative expenses and capital charge, to determine our target guaranty fee. Catastrophic loss was a specific number as well: it was the amount just above what our pricing model projected we could absorb on a pool of loans, based on the capital we put up and the guaranty fee we charged on it. (For a reformed Fannie and Freddie, the government would pick the stress scenario that determines this loss amount.) And an unexpected loss was any loss greater (or less) than the expected loss, but below a catastrophic loss (or above zero). Unexpected losses affected our profits, but did not threaten our solvency.

In the framework of a true risk-based capital standard the role of CRTs becomes straightforward. CRTs protect against unexpected loss, not catastrophic loss. The decision on using them therefore comes down to whether a guarantor wishes to pay the CRT interest costs in order to reduce the potential variability of its profits. That is a judgment the guarantor can and should make on its own; since CRTs do not directly affect solvency, there is no reason for a regulator to mandate their use. A regulator can make CRT use more attractive to a guarantor by giving equity capital credit for them, but in doing so it must be extremely careful in how it calculates their “equity equivalency.” If a regulator gives unwarranted equity capital relief for CRT issuance (as might easily happen under a number of current reform proposals) the guarantor will have a lower threshold of catastrophic loss, because it will have been allowed to exchange hard equity for less certain CRT coverage on too favorable terms, leaving some losses unaccounted for in a stress environment.

Finally, a true risk-based capital standard will make it much easier to use private reinsurance (rather than an explicit government guaranty or continued use of the current Treasury backstop commitment) to cover catastrophic mortgage loss, by providing specific thresholds, by risk category, at which that reinsurance kicks in. Again using as an example the capital and pricing grids published by FHFA in 2014, a reinsurer would know that for loans with LTVs between 61 and 80 percent and credit scores between 700 and 739, the companies could absorb nearly 7.5 percent in credit losses—through a combination of required capital and charged guaranty fees—before any claims had to be paid. Similar data would exist for all other risk categories. With a fixed-ratio capital standard reinsurers would know the size of a guarantor’s capital buffer but not the risk of the future business this capital will be protecting; to compensate for that uncertainty, they would have to add sizable cushions to their reinsurance quotes, if they didn’t elect instead to avoid the market for mortgage credit risk entirely.

In summary, then, the choice between a fixed ratio-based capital standard and a true risk-based capital standard for Fannie and Freddie, or any credit guarantor, is not a close call. The sole argument for the ratio-based standard—to create a “level playing field” with banks—is hollow, and a true risk-based standard offers overwhelming advantages in the areas of credit pricing, alignment of capital with risk, affordable housing, CRT evaluation and management, and the attraction of reinsurers such as property and casualty companies to the market for catastrophic mortgage risk.

The Moelis Blueprint is a practical and promising prescription for achieving the consensus objectives for a post-crisis secondary mortgage market. It uses proven mechanisms, can be accomplished administratively, and avoids the risk of transition to an untested system. One area in which I believe it should be refined, however, is the proposed approach to capital. The choice of a capital standard for Fannie and Freddie will have profound effects on how the companies conduct their business. With a true risk-based standard they will be able operate at much higher levels of effectiveness and efficiency than would be possible with a ratio-based standard. The more effective and efficient Fannie and Freddie are, the more valuable they will be to all of their stakeholders, and the easier and cheaper it will be to raise the amount of capital required to bring them out of conservatorship. Shareholders of the companies, the federal government and homebuyers all will benefit from the right choice on capital.