Notes from the Vault

Larry D. Wall, W. Scott Frame, and Lawrence J. White
March 2013

Is it possible to create a replacement for Fannie Mae and Freddie Mac that both offers government guarantees to investors and a truly fair deal to taxpayers? A recent report by the Bipartisan Policy Center (BPC) proposes to replace the two housing government-sponsored enterprises (GSEs) with new explicit government guarantees to investors against credit losses on mortgage-backed securities (MBS). The report calls for the government to take a "fourth loss" position behind private parties and to collect actuarially fair premiums. But would this be enough to provide a truly fair deal for taxpayers so they are fully compensated for the risks they bear?

Prior to the financial crisis, both Fannie and Freddie were quick to say that they posed no risk to taxpayers. However, since the crisis, virtually no independent observer would deny that taxpayers were always exposed to the potential for losses at these two GSEs. Since Fannie and Freddie's collapse in 2008, the BPC report notes they have taken $188 billion from the U.S. Treasury and are estimated to need cumulative draws of $191 billion to $209 billion by the end of 2015.

Given these losses, the housing finance proposals that require continued taxpayer guarantees also seek to reassure taxpayers that they will not suffer similar losses in the future. The taxpayer protection approach taken by the BPC proposal is similar to numerous other proposals from industry groups and some academic real estate economists (see the authors' review of these proposals). These proposals call for maintaining and possibly strengthening underwriting requirements, and perhaps also the representations and warranties by mortgage sellers, which have helped Fannie and Freddie to reduce their losses. Additionally, the BPC report and other proposals call for two additional mechanisms to protect taxpayers: some form of credit enhancement provided by a private firm, and actuarially fair guarantee fees (g-fees) paid to the government guarantor to cover its expected losses. The BPC report argues that under its proposal the government guarantee would apply only to "catastrophic risk."

In principle, it should be possible to establish a schedule of g-fees (contingent on the underwriting standards, reps and warranties, and credit enhancement) that would fairly compensate taxpayers for the tail risk of large losses on residential mortgages. But would setting actuarially fair g-fees be truly fair to taxpayers, and would such fees be sustainable in a political equilibrium?

Are actuarially fair g-fees truly fair to taxpayers?
The goal of actuarially fair fees is to set the g-fees so that over time their cumulative value equals the expected cost of the government's guarantee. That is, averaging across possible future outcomes, the government guarantee would not cost taxpayers anything. Nevertheless, guaranteeing MBS is risky to taxpayers because even if, on average, the program is actuarially fair, there are many possible scenarios in which the program would be extremely expensive to taxpayers, and the collected fees would not cover the actual losses. This constitutes a form of taxpayer provided risk capital, according to Boston College Professor Edward J. Kane—risk capital that by definition would not be compensated for bearing risk under an actuarially fair program.

Arguably, taxpayers do not require compensation for bearing risks that can be diversified away by the government. However, the risk of catastrophic losses in the housing sector is not a diversifiable risk. Such losses are highly correlated with the risk of a major economic downturn in which taxpayer income and wealth fall, government tax receipts drop, and requests for government social spending soar. In other words, the demands to honor the government guarantee on mortgage debt will almost surely be made at precisely the time when taxpayers and the government are least able to bear the cost of that guarantee.

Thus, a guarantee system that truly fairly compensates taxpayers should include a risk premium. Exactly how large a risk premium should be paid to taxpayers is beyond the scope of this discussion. However, as we shall soon see, even the goal of setting g-fees that are actuarially fair will be a daunting task.

Would "fair" fees be maintained in practice?
Although in theory there exists an actuarially fair g-fee and a somewhat higher g-fee that truly compensates taxpayers fairly for bearing risk, maintaining g-fees at either level is likely to prove almost impossible in practice. The difficulty is twofold: even good faith loss estimates by neutral experts are likely to exhibit wide variation, and political pressures will likely force fees toward the low end of these estimates (or they could be pushed even lower under continued political pressure).

The term "actuarially fair" is often applied in an insurance context, such as life insurance. Life insurers in the United States can base their estimates of the probability a person will die on observations of hundreds of millions of people. Moreover, each insurer is insuring tens of thousands of people. Thus, the random events that impact the mortality of any one individual, such as automobile accidents, are reasonably forecastable across the large group of insured individuals.

In contrast, what the g-fee is intended to cover is not the random risk that any individual will default on a loan but instead the risk that a large fraction of federally guaranteed mortgages will default, with losses that exceed the capacity of the private insurers and the guarantee fund built up from retained g-fees. If we return to the life insurance example, government guarantees of catastrophic residential mortgage losses are not like insuring the life of an individual. They are more like insuring against a pandemic that kills large numbers of people, such as the Spanish flu of 1918 that killed an estimated 20 million to 40 million people worldwide. Public health officials warn us such pandemics could happen again, for example, with a possible bird flu pandemic. However, the low frequency with which pandemics have occurred in modern times makes it difficult to forecast both the probability and likely death toll from a future pandemic. Moreover, the relevance of these historical data to future pandemics is somewhat questionable, given changes in transportation technology that can rapidly spread disease and advances in medical technology used to fight disease.

The same difficulties apply to estimating the expected present value of the catastrophic losses in the mortgage market. There are historical examples of such losses, but they happen only once every several decades. Also, the structure of the economy is likely to evolve over time so that even the recent drop in housing prices will lose some of its value in predicting the timing and magnitude of future catastrophic losses.

Thus, although the term "actuarially fair" can conjure up the impression of an objective, almost scientifically verifiable, measure with which to judge g-fees, the reality will be far different. Because such catastrophic events are so few, different experts using different data and methodologies can produce wildly different estimates of how frequently they are likely to occur and therefore wildly different estimates of the correct g-fee; there will be no objective way of determining which estimate is correct. As time passes, those seeking ever lower fees are likely to be able to support their case by pointing to years or even decades of no losses to the federal guarantor and to plausible expert estimates that "demonstrate" the g-fee is too high. That could lead to pleas to lower the g-fee (and/or the standards for underwriting loans or for private credit enhancers' financial condition) in the interest of facilitating wider homeownership. Certainly the precedent exists for lowering the cost of government guarantees, with FDIC insurance assessments dropping to zero for the 98 percent of deposits that were in the FDIC's "1A" assessment category on the eve of the recent financial crisis.

Will the political forces favoring lower g-fees and lower standards prevail? This brings us to arguably the biggest weakness of the BPC report and almost all other proposals for continued government involvement: their failure to address the reasons why Fannie Mae and Freddie Mac's precrisis federal supervisor, the Office of Federal Housing Enterprise Oversight (OFHEO), was not very effective. OFHEO was a weak supervisor because most of the politically active forces in the housing industry wanted it to be a weak supervisor (see James R. Hagerty's book on Fannie Mae; see also Gretchen Morgenson and Joshua Rosner's book on the financial crisis). Most of the housing industry gains from reducing the cost of government guarantees of mortgage lending and weakening the standards that are required to receive the guarantee. We see no reason to expect a change in the balance of political power in the future.

The bottom line
The BPC proposal and similar proposals likely would, over time, result in the same sort of subsidy being provided to mortgage finance that existed with Fannie Mae and Freddie Mac. The result is that taxpayers would be expected to make large payments to investors in mortgage-backed securities at a time when taxpayers' finances and that of the government are least able to bear these costs, and taxpayers would receive no compensation for bearing this risk. Further, odds are that whatever taxpayer protections are created at the founding of a government guarantor, they will be progressively weakened over time. Thus, it seems highly unlikely that such a guarantee program will result in taxpayers receiving a truly fair deal.

There are other reasons for supporting the BPC proposal or similar ones. For example, the authors discuss the concern that taxpayers simply cannot escape liability for a catastrophic decline in housing prices. In that case, we argue that explicit guarantees with some taxpayer protection may be better than implicit guarantees with no protection. But taxpayers should evaluate all proposals for continuing guarantees with their eyes wide open and do what they can to reduce the extent to which they are unwittingly exploited in the future (see Kane for some ideas).

Larry D. Wall is the director of the Center for Financial Innovation and Stability at the Atlanta Fed. W. Scott Frame is a professor of finance at the University of North Carolina–Charlotte and a visiting scholar at the Atlanta Fed. Lawrence J. White is Robert Kavesh Professor of Economics at New York University. The authors thank Ed Kane and Paula Tkac for helpful comments on the paper. The views expressed here are the authors' and not necessarily those of the Federal Reserve Bank of Atlanta or the Federal Reserve System.