Fannie Mae and Freddie Mac’s first quarter 2020 earnings were disappointing in one way but extremely promising in another. The disappointment came from the headline numbers, with Freddie reporting net income of only $147 million compared with $1.41 billion in the first quarter of 2019, and Fannie reporting net income of $461 million versus $4.37 billion in last year’s first quarter. Yet upon reading the details it became clear that each company’s results incorporated the effects of having adopted the current expected credit loss (CECL) method of reserving for loan losses, which shifted the basis for these reserves from losses already incurred to all expected future credit losses, and thus incorporated management’s current best estimates of credit losses from the Covid-19 pandemic. For both Fannie and Freddie, those estimates of loss were small by any standard.
In its earnings press release, Freddie said that the change from a $135 million (positive) benefit for credit losses in the first quarter of 2019 to a $1.233 billion (negative) provision for credit losses in this year’s first quarter was “primarily driven by higher expected credit losses on loans as a result of the pandemic, partially offset by the related expected recoveries from credit enhancements,” but it did not give dollar figures for these items nor did it specify the assumptions that produced them. Fannie did both. In its press release it said, “The company increased its allowance for loan losses to reflect the losses it currently expects to incur, including $4.1 billion as a result of the economic disruption caused by the COVID-19 outbreak, which are reflected in its $2.7 billion of credit-related expenses for the quarter.” It also gave two key assumptions behind this estimate in its 10Q filing with the SEC: that single-family loans in Covid-19 related forbearance would rise from their current estimated level of 7 percent of the total to 15 percent, and that 2020 home prices on a national basis would be flat compared with 2019.
During Fannie’s earnings conference call its CFO, Celeste Brown, disclosed more details of the assumptions behind the loss estimates, saying: “Our economists believe that the most plausible scenario is one in which the annualized Q1 GDP decline of 4.8% is followed by a more severe annualized decline of around 25% in the second quarter. Despite a forecasted rebound in the second half of the year, we expect full year 2020 GDP to fall approximately 3% and bounce back in 2021 to approximately 5% growth. We expect the unemployment rate to average 12% in the second quarter of 2020 with the peak of close to 15% before ending 2020 around 7%.” Fannie’s public filings and conference call did include qualifiers. In its 10Q it noted, “Our forecasts and expectations relating to the impact of the COVID-19 outbreak are subject to many uncertainties and may change, perhaps substantially, from our current forecasts and expectations,” while during the call Brown stated, “For the first quarter, the impact to our financials is largely based on management judgment, and by next quarter we will have more data points to consider and we’ll re-evaluate our economic outlook as well as our assumptions regarding forbearance and our loan loss allowance.”
Even with these caveats, both companies’ estimates of pandemic-related credit losses had to come as a welcome surprise and relief to policymakers, investors, analysts or any party interested in or affected by Fannie and Freddie’s financial health and ability to support the mortgage market during the potentially very challenging times ahead. Prior to the earnings reports, the prevailing discussions had been about whether the modest amounts of capital the companies had been permitted to retain since last June would be sufficient to cover their losses from the pandemic without requiring draws from Treasury. We now know from the companies themselves that, using the best estimates they’re able to make at the moment, they’ve reserved for all expected future credit losses without losing money in the quarter in which those reserves were added.
I have to say that my own sense is that Fannie (and by implication Freddie) is being overly optimistic in its assessment of how quickly and fully our country will be able to bounce back from the sudden shut-down of roughly one-third of the economy. Over 30 million people have filed for unemployment benefits in the last six weeks, and that number will grow in the coming weeks. Getting those people back to work won’t be nearly as easy as it was to let them go. But as everyone is saying, we’ll have to wait to see how things play out over the next few months. And in the meantime, we have a baseline for doing sensitivity analyses: the economic scenario Fannie’s management has laid out, and the credit losses the company expects will result from that.
Many commenters, it appears, had leapt to equate the pandemic with the 2008 financial crisis, and therefore based their fears of the magnitude of Fannie and Freddie’s potential credit losses on this prior experience. While that equivalence is false—the circumstances today are different in almost every way from what they were a dozen years ago—we nonetheless can use the 25 percent peak-to-trough home price decline during the crisis to create a worst-case loss scenario for our current situation. The $2.53 trillion book of single-family business Fannie held at December 31, 2007, just before the crisis began, now has twelve years of loss data, and it’s on track to experience a lifetime credit loss rate of 3.75 percent. An identical loss rate on Fannie’s $2.95 trillion single-family book at December 31, 2019 would produce lifetime credit losses of $111 billion, which, if they followed the pattern of the last cycle, would have nearly $80 billion hitting Fannie’s books before the end of 2024.
Could Fannie today be facing worst-case credit loss amounts anywhere near these levels? The answer is “no,” for two reasons. First, Fannie’s 2007 book contained a large number of loans with product types and risk features—such as interest-only adjustable-rate loans, no- and low-documentation fixed-rate mortgages and excessive risk layering—that contributed to roughly half of the credit losses on the 2007 book but no longer are permitted and thus barely are present in the 2019 book. Second, nearly half of the single-family mortgages in the December 2019 book are covered by some form of credit-risk sharing arrangement, over and above private mortgage insurance, which in a severe credit loss scenario will absorb a significant amount of these loans’ losses.
We’ll take these one at a time, starting with the higher credit quality. Fannie publishes periodic presentation packets on its Connecticut Avenue Securities (CAS) risk-transfer securities that include charts and tables adjusting Fannie’s historical credit performance by origination year for the differences in risk profile between the actual book of business then and the pool of loans it’s insuring with a particular CAS issue. While these data vary from packet to packet (because each CAS-insured pool is somewhat different), we can use their averages to reconstruct the losses of Fannie’s year-end 2007 book of single-family business had it been composed of mortgages with risk characteristics and features comparable to those of the mortgage pools Fannie has insured over the past half-dozen years. Doing this, the realized credit losses of the 2007 book fall to about 60 percent of what they actually were, making their lifetime credit losses not $111 billion but $67 billion, or $48 billion over the next five years (not $80 billion), before credit-risk transfers.
I’ve been a persistent critic of Fannie and Freddie’s credit-risk transfer programs, arguing that the companies were greatly overpaying for the right to transfer credit losses they were highly unlikely to incur. Up until late March, when social distancing took effect, that indeed seemed to be the case. For the past three years, the credit loss rate on Fannie’s post-2008 mortgages—which today account for 95 percent of its total book—has averaged a mere 2.0 basis points per year, and in 2019 it was only 1.7 basis points. Last year Fannie paid over $1.6 billion for risk transfers that generally don’t activate until cumulative losses as a percentage of the initial insured pool balance exceed 50 basis points (25 times the 3-year annual average of the post-2008 book), and they cover these pools for more than 400 basis points of cumulative losses (over 200 times the current 2.0 basis point annual loss rate). But now we have the pandemic, and in a worst-case scenario many of these risk transfer arrangements would pay off, to the great shock and dismay of the institutions and investors on the other side of them.
Over the past six and a half years, Fannie has been able to cover 46 percent of the single-family book of business it held at March 31, 2020 either with a CAS risk-transfer security (31 percent), a Credit Insurance Risk Transfer, or CIRT, arrangement with a mortgage insurer (10 percent), or a lender risk-sharing agreement (5 percent). Even in a worst-case scenario, however, not all of these arrangements will pay off, for three reasons. First, some risk-sharing is on older books that now have considerable seasoning and equity build-up, and won’t suffer large losses. Second, even in the highest-loss years some risk transfers will have coverage that exceeds the losses for that particular pool. And third, the most senior CAS tranches (called the M1), which typically cover losses in the range of 3.25 percent to 4.25 percent of the initial pool balance, can pay off very quickly if pool liquidations are high, and thus may not remain outstanding for long enough to absorb their share of losses.
The only way to account for all of the complexities in Fannie’s risk transfers and come up with a net worst-case credit loss number for the company’s December 31, 2019 book is to do projections of gross credit losses by origination year, then match those up with the CAS, CIRT and lender-risk sharing mechanisms associated with those books. I’ve done that using publicly available data. The result of this exercise, which obviously isn’t precise but also shouldn’t be too far off, is that in a repeat environment of a 25 percent nationwide decline in home prices as a result of the pandemic Fannie’s $67 billion in gross credit losses would be reduced by $30 billion because of loss transfers to third parties, leaving Fannie with $37 billion in retained losses, $27 billion of which would be recorded over the next five years.
Credit losses of $27 billion over five years is far from a catastrophic outcome for a worst-case scenario. As of March 31, Fannie had $6.0 billion in loss reserves against its post-2008 book of business, and another $7.2 billion reserved against individual loans from before 2009, most of which will become available to cover post-2008 losses as it gradually gets released. In addition, for the past three years Fannie has averaged $15.6 billion in pre-tax income before its provision for credit losses (which during that period has been an average benefit of $3.1 billion, because of releases of reserves from the pre-2009 book). Fannie’s existing collective loss reserve, likely releases from its pre-2009 reserves, and annual pre-provision income—coupled with the high quality of its post-2008 book of business and the credit risk transfers attached to that book—therefore should enable it to comfortably cover even a worst-case level of pandemic-related credit losses with no further draws from Treasury, as surprising as that may seem.
So, while we hope Fannie’s current estimate of $4.1 billion in pandemic-related losses turns out to be close to the mark, if it’s not it still seems highly likely that the actual outcome for its credit losses will be manageable. This means even in the worst case and during difficult times, Fannie and Freddie should be able to perform their intended roles of supporting the mortgage market, in spite of the fact that their March 31, 2020 core capital, at $13.9 billion and $9.5 billion, respectively, is far below a level anyone would deem adequate.
Which brings us to the impending FHFA capital rule. I wasn’t the only one to be thinking about how Fannie and Freddie’s business might be affected if, instead of the simple and straightforward (albeit outdated) capital standard the companies now have, they were to be subjected to what’s called the “conservatorship capital” standard FHFA promulgated in June of 2018, and is in the process of revising. Both companies addressed this in their 10Qs or conference calls, and made clear that the original FHFA standard has major flaws that if not corrected would have severe adverse impacts on their business in times like these.
In Fannie’s earnings conference call, CFO Brown noted that the inability to issue CRTs would raise Fannie’s required conservatorship capital, and added, “[i]f home prices decline, our capital requirements will increase due to the procyclicality in the current capital framework.” She also pointed out that under the June 2018 FHFA capital rule “loans in forbearance will carry a higher capital charge…credit risk capital increases by over five times with even one missed payment on a single-family loan.” This last feature means that if average capital for Fannie’s single-family loans was 2.0 percent, and 15 percent of these loans went into forbearance as is the company’s expectation, its capital requirement on all $3 trillion of its single-family loans held or guaranteed would jump to over 3.1 percent.
Freddie’s 10Q echoed the concern about CRTs, observing, “there may not be sufficient investor demand for CRT transactions at acceptable prices for the foreseeable future, and it is uncertain if there will be adequate demand for them over the longer term based on the potential impacts of the pandemic on mortgage performance.” And the Freddie 10Q also contained a table that starkly highlighted the interplay between two of the issues Brown raised: an inability to issue CRTs and procyclicality. This table showed that for loans covered by CRTs (at March 31, 51 percent of Freddie’s total), their conservatorship capital was 1.87 percent before CRTs but only 47 basis points after CRT credits. This post-CRT capital number looks too good to be true, and it is. FHFA’s June 2018 standard determines capital using current loan-to-value (LTV) ratios, and for Freddie those averaged 58 percent in March after a decade’s worth of home price appreciation had pushed them down from the original LTVs of 76 percent. I don’t have full historical data for Freddie, but during the financial crisis the current LTVs on Fannie’s loans rose from 55 percent at the end of 2005 to 79 percent six years later. If something similar were to happen to Freddie in a crisis environment, its pre-CRT conservatorship capital requirement would more than double, to over 4.0 percent, and with the CRT market shut down it would have no way to bring that percentage down at all for its new business, let alone to 47 basis points.
We can only hope that FHFA has scrutinized its June 2018 capital rule from the perspective of today’s circumstances as closely as Fannie and Freddie seem to have. Its original proposal was unnecessarily complex, excessively conservative, and had the great flaw of being too generous in good times and highly punitive in bad ones. As I said in my comment letter, FHFA can fix that proposal by making four basic changes to it: greatly simplifying it and removing the excessive conservatism; using original rather than current LTVs in the risk-based capital stress test; clarifying the roles of loss reserves, risk-based capital, minimum capital, excess capital and prompt corrective action, and not giving unwarranted value to credit risk transfers. On this last point, there is a role for CRTs in the revised capital standard, if investors will continue to buy them. But it’s not for FHFA to allow the companies to substitute contingent capital (CRTs) for upfront equity, which, as we now realize, will cause capital requirements and guaranty fees to spike during times of stress, when CRTs cannot be issued. Rather, Fannie and Freddie may wish to issue CRTs, at their discretion and based on the economics, to insure against the loss of equity in a stress environment, which is the role their CRTs are playing now.
So, here are some takeaways from Fannie’s and Freddie’s first quarter 2020 results. Most importantly, both companies believe that the high quality of their current books of business will enable them to make it through the rough times ahead, even with the modest amounts of capital they’ve been permitted to retain. Beyond that, analysis of prior stress experience strongly suggests that given their high-quality books, and the risk-transfer arrangements associated with them, even a 25 percent nationwide decline in home prices will not prevent them from continuing to carry out their missions with the capital and loss reserves they have today. It would, therefore, be the height of irony if what FHFA intends as an improved and more up-to-date capital standard actually impedes the companies’ ability to provide a comparable level of market support during whatever stress situation follows this one. FHFA must get this standard right. The pandemic has greatly increased the appreciation for the roles Fannie and Freddie play in the market—and strengthened the argument for their release from conservatorship—but it also almost certainly has delayed the timing of that release. Thus, if FHFA needs longer than the end of May to remove the idiosyncrasies and excessive conservatism from its June 2018 standard, it has that time, and should take it.