House under construction

Housing Perspectives

Research, trends, and perspective from the Harvard Joint Center for Housing Studies

The GSE Stress Test Results: Good News but Troubling Decisions

On August 13th the Federal Housing Finance Agency (FHFA), the regulator of the two government-sponsored enterprises (GSEs) Freddie Mac and Fannie Mae, released the results of its legally mandated Dodd-Frank Act Stress Tests for 2020 and 2021. Each of these, somewhat confusingly, reflects the results as of the end of the prior year, i.e. 2019 and 2020 respectively. 

The stress test as of a given date models earnings or losses over the following nine quarters based upon the Federal Reserve’s “severely adverse scenario.” Specifically, for year-end 2019 and 2020 results, this scenario assumes, respectively, that house prices decline by 28 and 23.5 percent, unemployment climbs to 10 and 10.75 percent, real GDP declines by 8.5 and 4 percent, and equity prices drop a full 50 and 55 percent. By comparison, during the financial crisis of 2008 – centered in a housing bubble built upon loose credit and high leverage – house prices fell about 25 percent. In other words, the stress test is (to use a slang term) very “stressy.” No one should believe it is manipulated to go easy on the financial institutions that undergo the tests; the reality is quite the opposite – it is truly “severely adverse.”

I see three major takeaways from these results.

The Good News

The first major takeaway is the good news that the announced results for Freddie Mac and Fannie Mae combined, when added to the history of the results since the first GSE stress test for 2013, paint a clear and simple picture: the GSEs have undergone an incredible amount of de-risking that should impress policymakers in housing finance.

Stress Test for Year Ending

Combined GSE Modeled Comprehensive Income (Loss)

  Without DTA Write-down With DTA Write-down
2013 ($90.3B) ($195.8B)
2014 ($73.4B) ($162.1B)
2015 ($51.3B) ($127.6B)
2016 ($35.1B) ($ 99.5B)
2017 ($41.9B) ($ 77.4B)
2018 ($18.0B) ($ 43.3B)
2019 ($ 7.18B) ($ 29.1B)
2020 $10.8B ($ 11.0B)

Percent Loss Reduction (2013 - 2020)

Over 100 percent 94 percent


That there are two sets of results – one ‘with’ and one ‘without’ deferred tax asset (DTA) write-down – reflects a rather complex tax accounting-related topic of whether such a write-off would occur given the modeled earnings, a full explanation of which is beyond the scope of this blog. The FHFA from its first publication of the results always showed both possible outcomes rather than opining on which accounting choice would likely apply. Key to the determination of whether DTA would or would not be written down is the likelihood of strong earnings in the following years. Given the large losses in 2013-15, my view is that “With DTA Write-down” is the outcome that would likely have occurred for those years. However, by 2018-20, the losses had become small enough that “Without DTA Write-down” for those years is likely to be the correct calculation. This means the risk reduction is even greater, with the nine-quarter modeled earnings going from a $195.8B loss in 2013 to a $10.8B profit in 2020.

There are three primary reasons for this dramatic decline in the modeled loss:

  1. House prices have increased strongly and steadily since 2013. Potential credit losses for a mortgage lender or guarantor are heavily tied to house price movements. Since year-end 2013, the FHFA’s seasonally adjusted index of house prices has gone up 55.5 percent through the end of 2020 (or 6.5 percent compounded annually) – a major increase. This dramatically reduces risk as borrowers have gained significant equity in their homes. Also, the largest underlying cause of this high level of price increase is a shortage of new home construction, which is not easily reversed; this is, of course, nothing like the excessively loose lending practices that drove prices so high in the 2000s until the bubble burst in 2007-8.
  2. The GSEs have actively taken steps to sell off risk, versus their pre-conservatorship business model of “buy and hold.” This includes the sale of impaired assets – such as modified or non-performing loans – as well as credit risk transfers on their performing book of guarantees. They pay away a share of their revenues to reduce risk in this manner, and it is working well.
  3. Since entering conservatorship, the GSEs have developed a credit risk appetite that is, in the spirit of Goldilocks, not too tight or too loose. Their performance in the pandemic shows that they are not being too loose, as demonstrated by their much lower rate of forbearances at 1.69 percent (as of the latest data) versus the combined FHA-VA at 3.95 percent or the private markets (meaning bank lending and private label securitizations) at 7.05 percent. The fact that the GSE market share was about 45 percent of the country’s single-family mortgage origination volume from 2014-19, and then grew to almost 60 percent upon the start of the pandemic (when private markets cut back), shows they are not being too tight.

In other words, when it comes to risk, things seem to be going rather well at the GSEs during conservatorship. Between taking proactive steps to shed risk, being balanced in their credit risk appetite, and benefitting from strong house price appreciation, the two firms are far less risky than anyone would have predicted when conservatorship began, while still faithfully pursuing their Congressionally-mandated mission.

Two Troubling Decisions

Two other major takeaways – specifically, related to decisions I find troubling – start with the curious fact that the 2020 results, reflecting year-end 2019, were not issued as scheduled last summer; instead, the FHFA doubled-up this year and released two years of results at the same time. This is not the norm. The FHFA is required to conduct the stress tests annually under the Dodd-Frank Act, which also requires  public disclosure of a summary of the results.

 As the pandemic began to significantly impact financial markets less than three months after the date of the 2020 stress test (i.e. year-end 2019), there arose a reasonable concern that large financial institution stress tests would be behind the curve in identifying a major pandemic-related increase in risk. As a result, after the Federal Reserve released its year-end 2019 stress tests for the largest banks on schedule, it then did an extra stress test as of mid-year 2020 to specifically provide transparency addressing this concern. The FHFA, by contrast, made a decision to not release the results of its year-end 2019 annual test (which had been completed by the GSEs and FHFA staff) as scheduled – the first of two troubling decisions. The logic disclosed by the FHFA for this decision was that the purposes of the Safety and Soundness Act would be "adversely affected" unless the results were not delayed until pandemic-related scenarios could be included. This logic made no sense to me then, nor has it since. It was the exact opposite of the action the Federal Reserve took (I note that the Federal Reserve is far more experienced in stress testing and regulation than the FHFA), and on top of that the FHFA never did release mid-year results including those pandemic-related scenarios.

What to make of this troubling decision? Simply put, it has terrible optics. It looks like the results were suppressed for political and ideological reasons. Specifically, at the usual time scheduled for the release (August 2020), then-Director Calabria was heavily involved in trying to complete a new regulatory capital requirement to apply to the two GSEs, which – as per his well-known viewpoint – he wanted to be very high. His narrative justifying such a very high capital requirement – namely, that the GSEs for years had excessively lax credit underwriting standards and thus would be subject to very large losses in a downturn – was directly contradicted and undermined by the stress test results from last year. Hence, one can reasonably conclude he may have used the pandemic as an excuse to suppress the test results because they contradicted his narrative.

Additionally, this episode certainly looks like a high-handed regulatory action, which sets a troubling precedent about regulator discretion. Since the Dodd-Frank stress test and the disclosure of a summary of the results is legally required, it certainly seems rather questionable.

Given that the stress test results from a year ago were known inside the FHFA in the summer of 2020, the second troubling decision was that the agency, just a few months later, went ahead anyway and finalized the proposed combined GSE capital requirement at a very high level – specifically $283 billion (based upon June 2020 data). This was generally regarded by commenters on the capital proposal as far too high. Given the actual stress test results from the end of 2019 (the last known at that time inside the FHFA, but not disclosed to the public until now), this means the required capital was 40 times the “without DTA write-down” amount of $7.1B (which I believe to be the likely outcome) and nearly 10 times the “with DTA write-down” amount of $29.1B (the unlikely outcome). Those are excessive ratios, reflecting that the regulatory capital requirement seems wholly inconsistent with the stress test results. And when considering the stress losses from the end of 2020, which even includes positive income for the “without DTA write-down” calculation, the regulatory capital requirement seems to be even more out of line with the reality of modeled stress loss.

To sum up, the doubled-up stress test results just released are good news for the US housing finance system because they show that the GSEs, as its biggest players, have very significantly reduced their risk exposure, making the entire system stronger and more stable. The results, however, are not good when it comes to evaluating the Calabria era at the FHFA, as they highlight two troubling decisions that appear to show politics and ideology unduly distorting proper regulation.