Current GSE Guarantee Fees Are Too Low to Be Consistent with Regulatory Capital: Does This Mean a Large Increase Is Coming?
The average guarantee fee (G-fee) of Freddie Mac and Fannie Mae, the two government-sponsored enterprises (GSEs), who currently finance about half of the nearly $13 trillion of outstanding first-lien single-family mortgages in the country,1 is among the most closely-watched numbers by housing finance policymakers and the mortgage lending industry.2 In November of last year, the Federal Housing Finance Agency (FHFA), the regulator and conservator of the two companies, issued its annual report on their G-fees (the G-fee Report), covering calendar year 2021.3 In that report, the FHFA disclosed that the average G-fee across all products was 0.46 percent.4 That average will change slightly each year to reflect the normal annual variation in the mix of products and borrower characteristics – for example, it was up by two basis points (bp) in 2021 over 2020, but down by the same amount from the prior year. Otherwise, the level has been strategically steady since it rose to its current range (between 0.45 percent to 0.49 percent) in 2014, after having been purposefully increased by the FHFA and the two GSEs in prior years. Unfortunately, that level began to be called into question by the adoption in 2020 of a new and much higher capital requirement by the FHFA, which in turn seems to require a much higher average G-fee. As such a higher G-fee has not yet been seen, it creates a major policy uncertainty overhanging the mortgage lending system – is a big G-fee increase inevitably coming?
How and why there is a problem
Why the average G-fee has been in the 0.45 percent to 0.49 percent range since 2014, rather than being materially lower or higher, does not seem to be well understood in the industry or among policy specialists. However, that is very understandable given its constantly changing historical rationale.
- Before conservatorship, it was set by the GSEs themselves at a low level as much for political reasons (in particular to help maintain their lobbying power in Congress, as further explained below) as economic ones.
- In the years immediately following conservatorship (which began in September 2008), the FHFA increasingly took over setting the average G-fee. Through 2013, the fee moved up strongly as part of the FHFA’s push to raise the cost of GSE mortgages in an attempt to “crowd in” more private market capital into mortgage lending.5
- From 2014 through 2019, however, the average G-fee increasingly became based upon a well-established financial markets concept:6 to generate, after covering expenses, a proper market return (also known as the “cost of capital”) on the capital that is required to support the risks being taken in the business.7 Both the capital requirement and the needed level of market return were determined by the FHFA, then under the directorship of Mel Watt8 (2014 to very early 2019).
This clear policy standard for setting the average G-fee was, however, thrown into some uncertainty in 2020. At that time, the FHFA – under its new director, Mark Calabria9 – revised the required capital to be a significantly higher number, but then did not commensurately increase G-fees. This has left the two GSEs earning since then a return well below a proper level, all according to FHFA calculations, on the revised regulator-set capital requirement.10 Was this perhaps indicative of a new concept for setting G-fees?
This question was in fact directly addressed in the most recent G-fee Report, where the FHFA stated that it did not change its concept. Instead, despite the adoption of the new and much higher official regulatory capital requirement in 2020, the agency just delayed the implementation of the higher capital level for two years. The FHFA then indicated that it ended this delay and began to employ the newer, higher capital requirement approach beginning in 2022.11 However, there is no evidence that there was a commensurate increase in G-fees in 2022, as such a large increase would have been a major political event in housing finance, with lots of loud complaining by the industry. Individual GSE public financial disclosures do, however, indicate there may have been a modest increase during 2022 (i.e., beyond just a year-over-year variance caused by mix changes), as discussed below.
Nevertheless, the two GSEs have for over a year now been officially (i.e., according to FHFA calculations) earning too low a return – much less than their cost of capital - on the single-family guarantee-and-securitization business, which provides their largest revenue stream. There are two key reasons why this discrepancy is not just an academic or arcane regulatory issue. One, the reality is that, since the two companies entered conservatorship, the provider of the capital that supports the risks being taken by the GSEs has been, for all practical purposes, the U.S. taxpaying public.12 This means that the taxpayers are now officially (again, using FHFA calculations) earning a sub-standard return on their investment, providing what amounts to a large, hidden, and never-congressionally-approved economic subsidy to the GSEs,13 which is not good public policy. Second, if and when the GSEs exit conservatorship and are re-privatized, it will be necessary for public market shareholders at that time (who would then be replacing U.S. taxpayers as the providers of capital to the GSEs) to believe they will earn a proper market return on their capital – but as estimated by themselves rather than the FHFA. Naturally, earning a significantly substandard return in the eyes of the public capital markets will delay or doom any such effort to take the two companies private, leaving them in a seemingly never-ending conservatorship.
Thus, the discrepancy between the current average G-fee and today’s regulatory capital requirement is a very visible source of policy uncertainty. Will the FHFA eliminate the inadequate return by dramatically raising G-fees? Will it revise the Calabria-era capital requirement – which was quite controversial when it was instituted because it was so much higher than its predecessor – to be considerably lower, i.e., something more in line with what was previously used? Will the FHFA somehow just reduce the level of market return required during conservatorship by adopting some new approach to calculating it?14 Or will the FHFA use a combination of two or more of the above-listed actions? And will any change be implemented quickly or perhaps incrementally over several years?
Impacting all this is the well-established politics surrounding the GSEs. Any major increase in the average G-fee will face very strong resistance from politically powerful housing industry groups - such as homebuilders, realtors, and mortgage companies, all of which benefit from mortgage credit being as low cost as possible. Housing advocates also would lobby heavily against such a large increase, as it makes it harder for a working-class household to become a first-time homeowner. Meanwhile, any reduction in capital requirements from today’s elevated levels would predictably be castigated by housing specialists from the political right and like-minded members of Congress, who will undoubtedly then accuse the FHFA of engaging in deliberate undercapitalization, such as existed pre-2008 (as explained below). Thus, the discrepancy might be allowed to continue for the time being to avoid the damned-if-you-do/damned-if-you-don’t politics associated with resolving it.
Given all this background, it is very understandable that where the G-fee and regulatory capital go from here is very hard to predict.
This article goes into the history behind this policy uncertainty, estimates what the impacts of each alternative alone to eliminate it would be, examines evidence that there may well have been a small increase in the average G-fee during 2022, and recommends that the best course of action, in my opinion, is for the FHFA to primarily rely upon revising the capital requirement formulae back to approximately what they were in 2019.
Additional G-fee history and background
Before conservatorship, as described above, the two GSEs set their own G-fees based on a combination of economics, market competition, and politics. In fact, based on my conversations with GSE and regulatory officials of that era, it is arguable that the G-fee was set more on political considerations than anything else.
The GSEs then were public-private hybrid organizations – private in terms of being shareholder-owned (having boards with fiduciary duties, seeking to earn a profit, and so on) but public in enjoying various special government-endowed privileges designed to help homeowners. However, the reality is that their design had its defects, allowing the two GSEs, once they became public shareholder-owned companies (Fannie Mae in 1968 and Freddie Mac in 1989), to exploit those privileges to produce large, never-approved-by-Congress subsidies that disproportionately benefitted their management and shareholders rather than homeowners. The main privilege so exploited was the “implied guarantee,” i.e., that the U.S. government, from the original privatization of Fannie Mae in 1968, officially disavowed formally supporting the GSEs in terms of U.S. budget calculations but intentionally set things up so that the broader financial marketplace believed the government had an obligation to never let the two companies default on their debt;15 this allowed the marketplace to largely ignore measures of GSE financial strength, in particular their capital ratios.16 The result was that the debt issued by the GSEs cost only a little more than what the Treasury pays itself, which is well under what even the most creditworthy traditional private sector large financial institution would incur. The two companies did not pay for the implied guarantee.
The GSEs were then each able to exploit this free implied guarantee to create a giant, profit-making subsidy for themselves. They raised unsecured funds, at those near-Treasury rates, to finance building up extensive discretionary mortgage-related investment portfolios. The two portfolios together peaked at over $1.5 trillion in 2016, which at the time was almost twice the size of the Federal Reserve’s balance sheet and became the largest source of GSE profits. Those profits were all due to the low cost of funding the portfolios, i.e., the benefit of the implied guarantee, rather than any superior investing prowess.17
The GSEs (supported by many other groups that benefitted from low-cost mortgage credit) also lobbied heavily and successfully to have very low regulatory capital requirements. This would have been somewhat self-defeating if investors in GSE debt instruments had in turn responded by demanding higher interest rates to reflect the higher risk of insolvency. But those investors did not do so because they looked instead to the implied guarantee for the creditworthiness of the two companies. There were two significant results of such low capital requirements: G-fees could be lower, generally benefitting homeowners;18 but also returns on equity were artificially raised, producing higher stock prices that benefitted shareholders and (via their stock-based bonus programs) executives.
Not surprisingly, the two companies highly prioritized protecting those subsidies, which meant maintaining political support in Congress to ensure no legislation was passed to take them away. As part of ensuring that support, the average G-fee at the time was thus kept low – in the 0.20 percent range – and generally did not vary by the borrower’s creditworthiness. This was a key pillar of keeping broad support across the various housing-related industries – realtors, homebuilders, and mortgage companies – who benefitted from low-cost mortgage credit. It also similarly kept the support of housing advocates, who liked not only low-cost mortgage credit but specifically that riskier borrowers (who tend to be working-class and first-time homebuyers) did not pay a rate higher than other, more creditworthy borrowers.19 This support was translated into extreme lobbying power in Congress, which ensured that the two companies did not lose the subsidies.20
In short, the GSEs treated the average G-fee as a type of “political loss-leader,” keeping it visibly low while most profits and high shareholder returns were made through the hidden subsidies of the implied guarantee-funded discretionary investment portfolios and the unduly low capital requirement.
This all ended in 2008, when even the most ardent GSE defenders in Congress went quiet as the two companies collapsed into conservatorship. Subsequent testimony and reports in Congress about the extent and effectiveness of GSE lobbying among elected members of the House and Senate were naturally very embarrassing to all involved.21 As part of conservatorship, and as a result of this sordid history, the GSEs were explicitly prohibited by Congress and the FHFA from conducting any lobbying.
Under conservatorship, the average G-fee then underwent dramatic change.22 It became increasingly risk-based over time (beginning shortly before conservatorship in early 2008 in response to the stresses of the global financial crisis, or GFC), as is typical for any private-sector financial institution. The GSEs and the FHFA together raised it to more than double23 its pre-conservatorship value and a more defined economic rationale was developed over time to determine its “right” level.24 (In addition, more recently, an increasingly complex system of cross-subsidies has been embedded in G-fees, intending not to change the average. The FHFA just recently announced the last such cross-subsidy changes in January 2023.25
The result, as described above, was that the average G-fee reached 0.48 percent in 2014 and then stayed in the 0.44 percent to 0.49 percent range for seven more years through 2021, as disclosed in each year’s annual FHFA G-fee report. (Evidence of the average G-fee’s movement in 2022, since the FHFA will not report official numbers until November of this year, is examined below.)
Solely increasing the average G-fee to achieve consistency with regulatory capital
In its latest G-fee Report, the FHFA explains how the capital system used from 2014 to 2019,26 called the Conservatorship Capital Framework (CCF), was replaced in 2020 by the much higher Enterprise Regulatory Capital Framework (ERCF). There was limited transparency about the capital required under the CCF, which is understandable since it was meant to be used solely inside conservatorship. There was incomplete transparency about ERCF until just this past year, when the GSEs first began to disclose it in published financial statements. Nevertheless, there is just enough historical information to reasonably estimate how much the G-fee will need to go up if the same return is to be earned on the ERCF as has been earned based on the CCF. Specifically:
- Freddie Mac's earnings press release for full year 2019 (but not since) disclosed that the CCF required, on average during that year, was $52 billion.27
- The FHFA, in the fact sheet accompanying its final rule amending the ERCF, stated that the September 30, 2021, ERCF for Freddie Mac was $123 billion.28
To compare the two, it is first necessary to adjust the 2019 CCF calculation to reflect the increase in the size of Freddie Mac’s guarantee portfolio two years later, specifically as of September 2021. From mid-2019 to the third quarter of 2021, Freddie Mac’s assets increased by 38 percent.29 I have therefore made the simplifying assumption that the capital required by the CCF would have had to increase by approximately the same 38 percent, i.e., to $72 billion, as of September 30, 2021, the exact date that the ERCF was disclosed to be $123 billion.30
Thus, just by switching from CCF to the ERCF, the capital required goes from an estimated $72 billion to $123 billion, or up by just over 70 percent. This is a very substantial increase and remains controversial to this day.
The impact on G-fees, if those are increased to be consistent with the ERCF, will be similarly substantial, based upon extrapolating the 70 percent company-wide increase in required capital to proportionately apply to the single-family business (which generates the vast majority of it). Of the current 0.46 percent average G-fee in 2021, about one-third represents items such as operating expenses and loan losses, i.e., things other than the return on capital, and the other two-thirds the needed capital return – or a split (with some rounding) of about 0.15 percent versus 0.30 percent.31 Increasing the 0.30 percent needed for a required return on capital by 70 percent means that it will need to rise to 0.51 percent, which I will round down to 0.50 percent for discussion purposes, or an increase of about 0.20 percentage points.32
Thus, to eliminate the current policy inconsistency, which emanates from the switch to the ERCF from CCF, solely by increasing the average G-fee would require it to go from 0.46 percent to about 0.66 percent (or, with the separate 0.10% excise tax that the government adds to G-fees included, from 0.56% to 0.76%).
Whether implemented all at once or via smaller changes over several years, such a major increase in the average G-fee will generate a tremendous negative political reaction from the mortgage banking, realtor, and homebuilding industries and housing advocates along the lines described previously. It would also run counter to other Biden administration housing-related policies, which seek to make housing more, and not less, affordable.
Solely reducing the capital requirement to be consistent with today’s average G-fee
Instead of solving the policy inconsistency solely by raising the average G-fee, as analyzed above, a clear alternative is to reduce the regulatory capital requirement. This would be, in essence, an unwinding of the ERCF and a return to approximately the CCF that was officially in place from 2017 to 2019 for setting the average G-fee. There are two questions associated with such an approach: how much would the capital requirement have to come down? and would a reduction in the required capital make sense, i.e., be “safe and sound” regulation and not akin to the politically biased undercapitalization that existed before 2008?
In terms of how much the capital requirement would have to come down, it would mathematically need to reverse the 70 percent increase calculated above. That means today’s ERCF-based capital requirement would have to come down by an estimated almost 60 percent (the inverse of going up by 70%) to get to the CCF level. This would make today’s average G-fee level produce a proper market return.
It is worth recounting that the CCF was developed, beginning in 2014, to be consistent with post-2008 concepts of proper financial institution risk-based capital requirements and then applied to the particular risks of the GSEs.33 It was originally meant to be used solely inside conservatorship – and so developed without any public transparency as merely an internal matter between the FHFA and the GSEs.34 Thus all the mechanisms for interest groups to influence such things (e.g., public comment periods) were bypassed. As far as I could discern then, it had no political bias one way or the other,35 generating much credibility with me and many others.
But does this large - about 60 percent - capital requirement reduction make sense?36 We know the original ERCF proposal was widely criticized for being too high. Of course, housing industry participants have a track record of opposing just about any increase in the cost of mortgage credit. But the public comments submitted about the ERCF to the FHFA went beyond that, with other voices that I view as more technical and independent also pointing out how unjustifiably large the increase was.37 But the most robust hard evidence that the ERCF is just too high – with $319 billion required across both companies, per the FHFA – is that this result is so wildly inconsistent with that of the government-mandated and -designed “severe adverse” stress test results, which recently showed at most just a $4.5 billion loss. Simply put, a large financial intermediary does not need to be capitalized at nearly 70 times38 its worst-case government-defined stress test loss.39
The lower CCF-based capital requirement would still equal more than 30 times the stress test loss, which certainly appears to me to be far from undercapitalization. In fact, undercapitalization is absolutely to be avoided. History shows us that before 2008 capitalization was done deliberately to “help homeownership,” but this ended up leaving the GSEs too weak to withstand the risks that turned into real losses during the GFC.40
Nevertheless, it has been most of a decade since the CCF was first developed. Instead of returning to it as it existed in 2019, it seems to make sense for the FHFA to update it as necessary, looking to be – Goldilocks style – neither politically biased low (as was true before 2008) nor high (which in my view is very much the case for the ERCF.)41 This would include, at the very least, adding a countercyclical buffer that was not included in CCF. Against the backdrop that the GSEs have been de-risking themselves generally for many years now, I expect the net result to be not that far off from the CCF level. In any case, it will be what it will be, and the average G-fee might need to be adjusted a bit higher or lower than what it is today.
Solely reducing the market return required to attain capital-vs.-average G-fee consistency
Coincident with developing the CCF as a good faith effort to come up with a credible, post-2008 capital requirement, the FHFA also had to develop what market return should be earned on that capital by the GSEs, i.e., their estimated cost of capital. If the two companies had stock that traded in a routine way (i.e., as if the two companies were not in conservatorship), this could be estimated to a high degree of accuracy from the history of the stock’s price and other financial metrics. However, in conservatorship, the equities of the GSEs don’t trade on each company’s earnings.42 So, instead, the FHFA selected a list of roughly comparable banking, insurance, and other financial firms, calculated the cost of capital for each, and averaged the results. This produced, at the time, about a nine percent market return requirement – and this was surprisingly stable throughout my tenure at Freddie Mac, which lasted till mid-2019.
Seemingly in reaction to the fact that a higher ERCF capital requirement might require significant G-fees raises, there began to orally circulate in housing policy circles on the political left a notion that the cost of capital should not be in the nine percent range, but instead around six percent. The justification that I heard was that since the GSEs were price-regulated like electric utilities while in conservatorship, and since a leading proposal for their possible exit from conservatorship would include continuing electric utility-style price regulation, then they should get a regulated electric utility level of cost of capital - and that this was about six percent. This number, interestingly, would leave the average G-fee roughly unchanged, i.e., the lower cost of capital would mostly offset the 70 percent increase in capital that resulted from switching from the CCF to the ERCF.
But is this argument valid? Or is it just a case, in a highly politicized industry, of a politically convenient advocacy rationale to justify not increasing G-fees? For several reasons, in my view, the argument is not substantially valid. The primary reason is that it does not seem to account for the extremely high earnings volatility of the mortgage business43 in contrast to the rather stable electric utility business – which I believe mainly drives the latter’s lower cost of capital rather than just the fact that it is price-regulated. However, as described above, if and when serious planning begins on conservatorship exit, it will be real investors who might purchase the shares of the two companies, which will set the cost of capital via what price they would be willing to pay at that time – and not what policy specialists, whether at the FHFA or elsewhere, say it should be.
What about the average G-fee in 2022?
The official FHFA G-fee Report for 2022 is due out in November. In the meantime, we can get a very good handle on what average G-fee will be reported at that time by looking at the public disclosures of the two GSEs. Averaging their quarterly reporting, the estimated G-fee by quarter is as follows44:
2022 quarter 1: 0.49%
Est. annual avg: 0.51%
On the surface, then, there seems to have indeed been a noticeable increase in the average G-fee of perhaps five bp (as calculated versus the 0.46% of 2021), which could be interpreted as being an intentional and partial implementation of about one-quarter of the total 20 bp increase previously estimated to be needed. But we know that the G-fee also moves each year in response to the mix of product and borrower characteristics – and there has lately been a profound change in that mix. Simply put, we have just gone from a market in which refinancings - which have significantly less risk than purchase loans, thus generating a lower G-fee on average - were a very high percentage of the mix (65% in 2021) to one where they are a comparatively low percentage (35% in 2022),45 reflecting the Federal Reserve’s dramatically-fast move from ultra-low interest rates to much higher ones. This product mix shift alone would have significantly increased the average G-fee. Thus, an initial conclusion that the average G-fee was intentionally raised by five bp, perhaps as the first move in a phased implementation of a 20 bp increase, no longer appears so certain – although perhaps the intentional increase was smaller, maybe something like two or three bp. Adding to the confusion is that the FHFA has been increasing the complexity of G-fees by implementing more extensive cross-subsidies, which are designed to not change the average – but which require an assumed mix for that calculation that may or may not prove accurate.
So, there is no hard evidence at this time, lacking an FHFA or GSE statement that it has intentionally increased the average G-fee, to know if a small increase has occurred or whether the rise to an estimated average G-fee in 2022 of 0.51 percent is instead solely or very significantly the result of extreme product mix (and cross-subsidy) “noise.”46
Where to go from here?
Housing finance is generally complex, and challenging issues often have complex solutions to match. But, in my view, that’s not so in this case. I see no need for a large G-fee increase at all. Instead, my recommendation for eliminating the policy uncertainty is based on my long-held conclusion that the ERCF is just too high, as strongly validated by its obvious inconsistency with the extremely low loss shown by the official stress test results. Thus, the path forward to resolve the current average G-fee inconsistency versus the ERCF is to revise the capital requirement down significantly by replacing it with an updated CCF. As a reminder, the latter was designed to be a good-faith effort for use inside conservatorship, based on the best post-2008 thinking on risk-based capital requirements, as applied to the specific risks of the GSEs, and without any intent to politically bias it high or low. I expect the result of using an updated CCF to look mostly similar to, although not exactly like, the CCF that had last been used in 2019.47 This would eliminate the inconsistency and, therefore, the need for a major increase in today’s average G-fee, although a small increase or decrease certainly might result. To me, this recommendation builds upon the best financial thinking of how much capital is needed for a financial institution, reflects safe and sound prudential regulation of the GSEs, and is not politically biased.48
Nevertheless, this straightforward solution has not occurred, and it is not clear why. Perhaps the Biden administration49 does not want to expend the time, hassle, and political capital needed to do such a revision, judging other issues it faces to be more important and urgent. Or maybe it believes that, in conservatorship, a full return on the risk just doesn’t need to be earned, even though this implicitly means there are big implicit taxpayer subsidies going to the two GSEs.
In the meantime, we will just have to “watch this space” to see how it all evolves.
Is this much ado about not much?
The astute reader may notice that the amounts discussed in this paper represent quite modest costs to a typical borrower. With today’s mortgage interest rates at over six percent, an average G-fee of 0.46 percent is small to start with (under one-tenth of the total). Being very concerned about increases of perhaps 20 bp (i.e., one-fifth of a percent) or maybe just five bp (i.e., one-twentieth of a percent) seems a bit questionable when the mortgage rate has gone up by a full three percentage points and more in just the last year. In fact, the mortgage rate today can easily go up or down by five bp or even 20 bp in a week or two because of routine market interest rate moves.
For the typical borrower of a 30-year fixed-rate mortgage with a $250,000 principal balance50 and carrying a six percent interest rate, the monthly payment would be $1,498.60 per month. If the rate were instead 6.20 percent (i.e., adding the full 20 bp needed to provide immediate consistency with the ERCF), the monthly payment would be a modest two percent higher, equal to $32.30 more per month. (A five bp increase, i.e., a partial implementation of the 20 bp increase, would only raise the payment by about ½ percent , or $8.05 per month, an almost negligible amount.) This does not seem to me to be very significant when it comes to affordability, and it is unclear whether it would have any measurable impact on the aggregate volume of home purchases. Nevertheless, I note that housing advocates would disagree, as in my experience, they ascribe large impacts to small changes in mortgage interest rates and thus are very dedicated to keeping a strong focus on any potential change – this is one reason why the average G-fee is so carefully scrutinized
However, financial intermediaries generally have profit margins tied to the “spread” – i.e., the difference between the interest rate earned from their borrowers versus what needs to be paid to their funding providers. In this case, that means that primary market lenders, a very important constituency in the U.S. mortgage financing system, would be heavily impacted by such a 20 bp increase (or even a five bp one) as it is a high percentage of their spread.51 So, a 20 bp increase in the average G-fee represents, to a primary lender, a cost increase that would put very significant pressure on their profit margins, depending on what percentage of the increase will be passed through to borrowers at any particular point during the economic cycle. Thus, primary market lender industry associations – there are several of them - would naturally heavily fight any such increase.
Of course, getting the needed consistency between the average G-fee and the regulatory capital requirement is also quite important to the GSEs themselves, as their spread is mainly the G-fee itself.52 Right now, their earning a considerably substandard return is counter to any proper notion of safety and soundness (although this can be less obvious during conservatorship); it also means there is a large, covert taxpayer subsidy that, as described above, is bad public policy. Solving the problem by revising the capital requirement downward from today’s ERCF, as long as the revised level is properly constructed (which the CCF mostly was, in my opinion), is also consistent with the safety and soundness of the GSEs.
Thus, the potential change in the G-fee related to the ERCF is indeed much ado about not that much for the typical homeowner, as the amounts and percentage increases involved are relatively small. But when it comes to the primary mortgage lending industry and the two GSEs themselves, where the interest rate spread that generates their revenues is so much smaller, how the policy inconsistency between the ERCF and the average G-fee is resolved is very much a consequential decision that requires focus and attention to have a safe and sound, efficient and properly operating mortgage financing system.
 See “Housing Finance at a Glance,” December 2022, from the Urban Institute, page 6, which reports the market as being $13.2 trillion in size. Excluding the $0.4 trillion of home equity loans (i.e., second-lien borrowings) leaves $12.8 trillion for the first-lien amount.
 For an explanation of why it is so closely watched, see the Addendum at the end of this document.
 “Fannie Mae and Freddie Mac Single-family Guarantee Fees in 2021,” November 2022, page 2. https://www.fhfa.gov/AboutUs/Reports/ReportDocuments/GFee-Report-2021.pdf. See also the predecessor reports for previous years, which are issued by the FHFA in the legally mandated annual report to Congress.
 The FHFA actually reported it to be 0.56 percent, but 0.10 percentage points of that total represent a separate excise (i.e., not income) tax that is collected by the GSEs on their loan purchases and remitted straight to Treasury. So, while nominally the average G-fee was 0.56 percent, to the GSEs economically it was actually 0.46 percent, which is what they received to cover their operating costs, losses, income taxes, and a return on the capital that supported the risk they were taking. In this paper, I will utilize this economic G-fee, i.e., 0.46 percent, unless otherwise specified.
 This was the policy of the FHFA when it was under the control of Acting Director Edward Demarco (2009 to very early 2014). See, for example, the press release, “FHFA Announces Increase in Guarantee Fees” (August 31, 2012), where the first paragraph says, “The changes to g-fee pricing represent a step toward encouraging greater participation in the mortgage market by private firms…” https://www.fhfa.gov/Media/PublicAffairs/Pages/FHFA-Announces-Increase-in-Guarantee-Fees.aspx. Also, the 0.10 percent excise tax set by Congress became effective in April 2012 – so primary mortgage lenders saw an increase in the G-fee even if the two GSEs did not receive any more revenue, as previously explained.
 It was officially adopted in 2017.
 The FHFA G-fee Report describes this using slightly different language: “Guarantee fees are intended to cover the expected credit losses, administrative costs, and the cost of capital associated with guaranteeing single-family loans.” The cost of capital is sometimes defined as the level of return needed for the market to value a company’s shares at book value (an accounting measure of the shareholders’ investment in a company). Of note, this is very consistent with how rate regulation historically works for electric utilities.
 The needed level of market return, or cost of capital, of a company is normally calculated based on the history of its share price and other financial metrics. However, in this case, it was calculated by the FHFA based on the average cost of capital for a range of other large banking, insurance, and financial companies as a proxy since the underlying shares of the two GSEs do not trade, in conservatorship, based upon economic fundamentals.
 Mark Calabria was FHFA director from April 2019 to June 2021.
 In conservatorship, the GSEs do not yet have this full amount of required capital. It would presumably be required to be in place for any exit from conservatorship to take place.
 See G-fee Report, page 6, footnotes 12 and 13. It is worth mentioning that this would apply to all the risk-versus-reward decisions at the two GSEs, not just the setting of G-fees.
 This reflects that the creditworthiness of the GSEs was re-established in conservatorship by Treasury entering into a support agreement with the two companies, called the preferred stock purchase agreement (PSPA). As that agreement has played out through various amendments, Treasury – acting as a channel for the taxpaying public – stands behind GSE obligations and effectively owns almost 100 percent of the earnings of the two companies.
 My rough estimate of that subsidy would be about $13 billion per year. The calculation behind this will be described further below.
 Of course, this would only eliminate the inconsistency until serious conservatorship exit discussions get underway - when return-on-capital requirements would be set by potential stockholders, rather than based upon an FHFA-chosen formula.
 For example, the GSEs had special direct lines of credit to the U.S. Department of the Treasury, which was taken by the market as a signal of support. But, at its core, the implied guarantee was based on the fact that the role played by the GSEs was so pivotal to the economy that the government could not take the pain of their failure. While there were subsequent legal requirements for the government to officially disavow any guarantee of GSE debt, in 2008 the implied guarantee was shown to be very real, with the conservatorship and PSPA being operated to make the creditors of the two companies whole (although shareholders were not). Interestingly, the credit rating agencies (e.g., Moody’s, Standard & Poor’s) stated clearly in their credit rating publications of the time that the AAA given to the two GSEs was based on that implied guarantee, just as they do today (in technical language) for other organizations which still enjoy such an implied guarantee (e.g., the Federal Home Loan Bank system).
 Thus, only the government had a strong interest in ensuring the financial strength of the GSEs. Unfortunately, it did not do so prior to conservatorship. During much of conservatorship (specifically 2012 to 2019), strangely, capital was depleted by design. However, beginning in late 2019, capital began to be accumulated and currently stands at $97 billion for the two firms.
 When I became CEO of Freddie Mac, I had an analysis of the historic investment portfolio profits conducted and this was the clear conclusion.
 It is always important to remember that the GSEs do not deal directly with homeowning borrowers. Instead, they purchase loans originated by primary market lenders. Thus, lower G-fees reduce cost to those primary market lenders, and the expectation was and still is that most of it will, via competition among those lenders, be passed through to actual homeowners.
 This phenomenon is known as cross-subsidization through the G-fee, where the average stays unchanged, but some borrowers pay too high a rate given their low risk, and others pay too low a rate given their high risk. The FHFA is currently engaged in increasing cross-subsidization.
 A great example of this extreme lobbying power occurred during the presidency of George W. Bush, when the administration and the Federal Reserve proposed to limit the size of the investment portfolios of the two companies in 2005. See https://www.everycrsreport.com/reports/RS22307.html. The GSEs won that battle, and no limits were imposed until conservatorship, when they were included in the original PSPA in September 2008.
 In particular, see the report of the Financial Crisis Inquiry Commission, February 2011, http://fcic.law.stanford.edu/report, and also specifically the testimony of Armando Falcon, who had been the head of an FHFA-predecessor regulatory agency (the Office of Federal Housing Enterprise Oversight or OFHEO) https://ypfsresourcelibrary.blob.core.windows.net/fcic/fcic-testimony/2010-0409-Falcon.pdf.
 See the G-fee Report, pp. 12 and 13, for a summary of G-fee actions over the years.
 And this excludes the impact of the 10 basis point excise tax added. Including that, they almost tripled.
 See the G-fee Report of 2013, issued in November 2014. https://www.fhfa.gov/AboutUs/Reports/ReportDocuments/GFeeReport1120914.pdf.
 See “FHFA Announces Updates to the Enterprises’ Single-Family Pricing Framework,” https://www.fhfa.gov/Media/PublicAffairs/Pages/FHFA-Announces-Updates-to-Enterprises-SF-Pricing-Framework.aspx.
 It was only officially adopted in 2017, but in practical terms began to be partially used as far back as 2014.
 See Freddie Mac 2019 Full-Year Earnings press release, page 6, https://www.freddiemac.com/investors/financials/pdf/2019er-4q19_release.pdf. As CEO of Freddie Mac, I emphasized such disclosure; I retired on June 30 of that year.
 See chart on page 3 of the Fact Sheet. https://www.fhfa.gov/Media/PublicAffairs/Documents/Fact-Sheet-Final-Rule-CRT_2252022.pdf.
 This is an unusually high growth rate, caused by the economic distortions of the pandemic.
 This assumes the riskiness of the mortgage portfolio did not change during this period, which is one source of possible inaccuracy in the estimate. The riskiness, in fact, likely declined as house prices increased so much during the pandemic – which would have made the apples-to-apples comparison of the two capital systems even more extreme, i.e., showing an increase greater than about 70 percent.
 During my time at Freddie Mac, the non-capital expenses amounted to about 0.15 percent to cover both operating expenses (such as salaries, data processing costs, etc.) of 0.10 percent to 0.12 percent and also credit losses of roughly another 0.04 percent. With the large growth of assets during the pandemic and the run-up in home prices, using these historic levels may be a small overestimate.
 Because it has not been implemented, and assuming the ERFC is the proper capital requirement, this 0.20 percent additional G-fee thus also measures how much the taxpayers who support the GSEs are being short-changed. When applied to the approximately $6.5 trillion of outstanding single-family guarantees, this produces a $13 billion economic subsidy from taxpayers to the GSEs.
 Those post-2008 concepts were developed by banking regulators.
 The CCF was later used, in 2018, as the basis for an initial proposal for a formal regulatory capital requirement, which included going out for public comment. However, it was never finalized as Director Calabria supplanted it with the ERCF.
 For full disclosure, I was a major participant in its development, especially in the earlier years of 2014 and 2015, and so am speaking from first-hand observation.
 The FHFA is now engaged in a second round of revisions of the ERCF under its latest director, Sandra Thompson. The revisions are, however, not aimed at materially changing the level of capital required, but just the specifics of individual items. This is being done so that those specifics create incentives for the two GSEs to engage in proper, and not distorted, risk-versus-reward decision-making at the transaction level.
 In particular, it was pointed out that the ERCF relied very heavily on multiple judgmental buffers to raise the total required, as well as having various “floors” that override the economics of what capital would be required to absorb certain risks, which also had the effect of raising the total.
 The ERCF at $319 billion divided by the $4.5 billion loss equals 70.9 times. This covers both GSEs.
 I have previously written on this topic, concluding that the ERCF is well too high (more than double) what is needed to be consistent with the latest stress test results. See “The Latest GSE Stress Test Results: Showcasing the Need for Regulatory Capital Revision (part 2), August 2022. https://furmancenter.org/thestoop/entry/the-latest-gse-stress-test-results-showcasing-the-need-for-regulatory-capital-revision.
 It is worth noting that overcapitalization, in the interest of being conservative, is not without its downside. In effect, it acts as an unnecessary tax on homeownership, which is bad public policy.
 Director Calabria had long advocated for a shrinkage of the GSEs, and made it clear in private and public comments while director that this was still an objective he had. By developing an unduly high capital requirement, the GSEs would be less competitive, aiding such market share shrinkage. This, to me, makes the ERCF politically biased high as much as the requirement prior to conservatorship was politically biased low. (Given the nature of politics, all involved will naturally deny any such bias.)
 Instead, they seem to trade mostly on the small possibility that lawsuits against the government will result in large gains to the shareholders. To date, the shareholders have lost all such lawsuits.
 As a recent example of this volatility, consider that Fannie Mae’s Q IV 2022 net income of $1.4 billion was down 42 percent from the prior quarter and a very large 73 percent from Q IV of the previous year. See their press release. https://www.fanniemae.com/media/46286/display.
 See the quarterly earnings press releases of Fannie Mae and Freddie Mac, available on their websites. I have simply averaged between the two companies.
 These figures are for Freddie Mac. See its annual report Form 10-K, page 35. https://www.freddiemac.com/investors/financials/pdf/10k_022223.pdf. Fannie Mae will have similar figures.
 The quarterly pattern of the G-fee in 2020 validates this conclusion. After rising to 0.53 percent in Q3, it dropped significantly back to 0.50 percent in Q4. That kind of quarterly volatility reflects, at the least, significant mix and/or cross-subsidy issues, and not solely a permanently higher G-fee.
 The biggest change would probably be the addition of a counter-cyclical buffer.
 I note that the FHFA just announced a proposal for some technical and targeted changes to the ERCF, but they do not produce a material change in the total amount of capital required.
 Any change in the capital requirement would require both FHFA and Treasury approval, as a Calabria-era amendment to the PSPA gave Treasury such authority. Thus, it really would be a Biden administration decision, not a technical regulatory one produced by the FHFA.
 Given that the focus on keeping G-fees affordable should be aimed at the first-time homebuyer and others with more marginal creditworthiness, this level of mortgage amount is reflective of the appropriate policy focus.
 For primary market lenders, the spread is between what interest rate (including fees) they can charge homeowners versus what is charged by the GSEs to purchase loans from them (or by other mechanisms for funding their loans). One measure of the spread is the “net production income” reported by the industry – and historically, it has averaged only 0.60 percent. (See my article “The Policymaking Implications of Record-High Mortgage Origination Profits During the Pandemic. https://www.jchs.harvard.edu/blog/policymaking-implications-record-high-mortgage-origination-profits-during-pandemic.) Thus, a potential 0.20 percent increase, when it is unclear how much can be passed through to borrowers, is a major event to primary market lenders.
 The underlying rate to which the G-fee is added, known as the “base rate,” is mostly passed through to the primary lender. Thus, the G-fee is functionally the spread between the rate they earn and what they pay.