Government Mortgage Interest Rates: A Serious Discussion about the Intertwined Topics of Risk Adjustment and Cross-subsidies

Research & Policy | May 31st 2023 | Donald H. Layton

The Federal Housing Finance Agency (FHFA), the regulator and conservator of Freddie Mac and Fannie Mae, the two government-sponsored enterprises (GSEs), has been very prominent in the news lately.  This is due to the controversy surrounding how the agency has changed the specific amounts by which the two GSEs risk-adjust the interest rates at which they purchase mortgage loans. This has unexpectedly led to heated public commentary, including two highly critical Wall Street Journal editorials. The core claim made by critics is that the changes, implemented on May 1, are a mechanism for a new and major economic cross-subsidy from less risky borrowers to more risky ones. Such a mechanism, they allege, was created by the Biden administration to redistribute wealth and income in a surreptitious manner.  The FHFA, meanwhile, has stated that there is no intention to do such redistribution and that the changes resulted instead from updating the GSE risk adjustments, mostly set about a decade ago, to be consistent with the rule it adopted in 2020 for minimum required capital. 

The purpose of this article is not to wade into that firestorm but to use its occurrence as a springboard to address the very serious  – and not broadly understood or openly discussed  – interrelated topics of interest rate risk adjustment (or the lack thereof) and the embedded and often hidden cross-subsidies that today are found at the four government mortgage agencies, i.e., the two GSES plus the Federal Housing Administration (FHA) and the U.S. Department of Veterans Affairs (VA).1    My view is that the current way they operate – which varies considerably between the four agencies – would be significantly improved if they all applied three key principles: 

  • Do risk-adjusting on a practical and easily implementable basis, very much enabled by today’s information processing capabilities.
  • Separately and directly target the borrowers that policymakers decide should be specifically subsidized by a below-market interest rate, and just as overtly determine how the cost of such subsidies will be paid for. 
  • Provide transparency to policymakers and the general public about how it all works, including information on all the cross-subsidies involved and the accuracy of the risk-adjusting.

Necessary Background

The four government mortgage loan agencies dominate America’s first-lien mortgage market by generating about 70 percent of the funding of the $13 trillion market.2  In simple terms, the “retail” interest rate most homeowners pay on their mortgage is based on primary market lenders adding a markup to the “wholesale” interest rate they pay on funding obtained through one of those four agencies. Over the near century of evolution of this dominant government role, the business model used by all four agencies has come to rely on packaging those loans into mortgage-backed securities (MBS) and then guaranteeing the investors in the MBS against credit losses on the underlying mortgage loans.  As a result, the wholesale interest rate equals whatever the global capital markets require on the related MBS plus a fee levied by those agencies to cover the cost of that credit guarantee and their other expenses.3

This business model is very complex.  Adding to the complexity is that those agency fees4 will vary for different reasons.  Specifically, those fees today vary to reflect: (1) how much the particular agency adjusts its fee to fully or partially reflect its estimate of the riskiness of each mortgage loan, (2) how much some borrowers are subsidized with under-market interest rates, including due to qualifying for programs specifically designed to benefit low- and moderate-income (LMI) or other specifically chosen categories of borrowers, and (3) how much other borrowers pay above-market interest rates to fund those subsidies. The three categories are often talked about within a framework of the agencies running cross-subsidies, i.e., where some borrowers pay more than is economically justified and others pay less, all approximately adding up to a zero-sum result.  This is in addition to the fee sometimes varying by product type (e.g., a purchase vs. refinancing loan) or amount.  The agency fees can take the form of an upfront fee, a per annum fee, or a mix of both.5

In terms of the design of the government mortgage agencies, there is a general hierarchy where each of the agencies has its intended focus and resulting risk profile, albeit with overlaps:

  • The GSEs.  The two companies are aimed at broad middle and upper-middle-class borrowers.  They have the least risky portfolio, as exemplified by the current serious delinquency rate6 of just 0.62 (Freddie Mac) and 0.65 percent (Fannie Mae).  They are structured as stockholder-owned companies (even while they remain in long-term conservatorship) which need to be properly capitalized, like any large financial institution, and then to earn a market return on that capital.7 Their guarantee, given to MBS investors against credit loss, is backed by the government via a support agreement between Treasury and each GSE, which has certain limitations.8
  • The FHA.  This agency’s role has evolved over many years. In the last several decades, it has been primarily aimed at riskier mortgage borrowers, i.e., those who need the extra help from the FHA’s very low funding costs (see below) in order to get over the finish line to become acceptably creditworthy. Its current serious delinquency rate is 4.86 percent, about eight times that of the GSEs.9 The FHA is a direct unit of the U.S. government and is thus not required to be capitalized like a private sector company, thereby relieving it of any burden to generate a return on capital;10 it is also authorized to use (done via GNMA) the full faith and credit of the U.S. government to support its guarantee against credit loss to MBS investors.  These two features, and a third related to tax, result in its funding costs being especially low – I estimate it to be at least 0.50 percent per annum below that of the two GSEs.11
  • The VA.  This agency’s mortgages are offered on very favorable terms as a way to deliver a special benefit to veterans and active military personnel.  The riskiness of its mortgage credit exposure has long been greater than found at the GSEs but less than found at the FHA – its current serious delinquency rate at 2.43 percent is about four times that of the GSEs but about half that of the FHA.  VA mortgages enjoy about the same low funding costs as the FHA since the VA is also a direct unit of the government with no need for capital or earning a return on it.  Investors in the related MBS are guaranteed against credit losses by GNMA and are similarly backed by the full faith and credit of the government.

It is important to clearly state that the analysis in this article is based on the economic reality that if a loan does not have an interest rate that is appropriately risk-adjusted for its creditworthiness, then by definition there will be cross-subsidies in operation, even if that is not overtly stated. Specifically, those paying more than is economically justified, thereby generating excess returns, are a source of subsidy dollars; those paying less than economically justified will be receiving a subsidy.  As background, in the private sector, all financial institutions work under the principle that every loan should have an interest rate that reflects its estimated12 creditworthiness; therefore, such cross-subsidies are relatively rare.13  For example, it’s likely that if any bank tried to do otherwise and routinely charged “flat pricing” (i.e., every borrower paying the same interest rate), its regulators would cite it for engaging in an unsafe and unsound practice.14  Only when government gets involved –  which is when policy and political concerns are combined with advocacy by ideological and economic interest groups – does one sometimes see something different, with the inevitably resulting cross-subsidies.  As will be described below, this is happening to a significant degree today in the American mortgage system, with a distinct lack of transparency as to who is overpaying and thus providing those subsidies versus who is underpaying and so benefitting from them. 

Why Now? – The Current Political Firestorm about GSE G-fee Changes

The FHFA implemented on May 1 a previously announced change in the GSE “pricing grids,” i.e., the level of the G-fee as it varies for different levels of creditworthiness.  The grids are structured to show the specific G-fee that would apply, for any class of products, as measured by a combination of credit score and the loan-to-value (LTV) ratio, all organized via a 9X9 grid into 81 “cells.”15  

In late April, reactions to the new GSE G-fee grids – which until then had been the object of discussion mainly within the industry and housing policy community – broke out into the realm of cable news, social media, and politics.  Conservatives were very vociferous in opposing the changes, most notably in two prominent Wall Street Journal editorials: “Upside Down Mortgage Policy” (April 22, 2023) and “Spinning Federal Mortgage Fees” (April 28, 2023). The changes then became the subject of hearings and legislation in the Republican-controlled House of Representatives to reverse the changes.16 Lawsuits by state-level Republican officials were also threatened. In short, this was a classic media-political firestorm about a relatively technical topic that is normally of interest only to those inside the mortgage system.

The accusations, although not always precisely stated, have at their core one of three interrelated claims. 

  • The most extreme criticism is that the new G-fee grids were implementing an “upside-down” credit risk charging scheme, i.e., charging most or all low-risk borrowers a high G-fee (and thus interest rate) so that the GSEs could then charge most or all high-risk borrowers a low one. This was described on the one hand as unfair, since it relied on overcharging low-risk borrowers “who had played by all the rules” and, on the other hand, as unduly incenting bad loans at the GSEs (by charging too little for high-risk loans) in a quasi-replay of the lead up to the mortgage bubble of 2005 to 2008.  This claim is simply untrue.  The second of two articles from the Urban Institute showed quite clearly, and I believe accurately, that the GSE G-fee grids remain distinctly “right side up,” i.e., having higher interest rates for higher risk.17
  • A less extreme version of the criticism is that the changed G-fee grid, while it still maintained a traditional “right side up” approach where higher risk loans were charged higher G-fees, significantly flattened the grid. This means that the FHFA incrementally increased the G-fee on most low-risk loans and reduced it on most high-risk loans. Critics also claimed that the Biden administration did this as some sort of stealth income or wealth redistribution program. Again, this was criticized for unfairly penalizing low-risk borrowers.  In fact, it is absolutely true that the flattening did occur for most, albeit not all, cells in the pricing grid.  The FHFA, though, stated that this was solely the result of updating risk estimates mostly made over a decade earlier to be consistent with the regulatory capital requirement adopted by the FHFA in 2020.  So, at this point, the controversy is mainly about the motivation for the changes. 
  • The third criticism is that government mortgage agency programs designed to help LMI borrowers18 – especially at the GSEs – are rife with credit underwriting problems and that the announced G-fee changes would add even more high-risk loans with inadequate revenue to cover their risks, all inevitably leading to a replay of the subprime crisis. In my view, this criticism is not valid.  While the risk of such loans, on average, is undoubtedly somewhat higher than the overall average among those financed by the GSEs, their risk also appears to be within the usual range of acceptable creditworthiness (referred to as the “credit box”) of the GSEs.  This will be further discussed below. 

On May 15, the FHFA issued a request for information (RFI) on the pricing grids.19  This was obviously in reaction to the criticisms.  The RFI document20 itself has given the FHFA an opportunity to explain what it did and why and to educate the public on various aspects of the topic. It also asked for input on a wide range of questions.  Comments are due in ninety days. 

The FHA – Minimal Risk Adjustment; Cross-subsidies Should Be Better Targeted

The FHA’s business model is to insure mortgages that are then issued via GNMA.  Its market share has roughly averaged 10 percent over the last several decades, with considerable variation through the economic cycles. 

FHA Interest Rate Adjusting.  The FHA does minimal adjusting of its interest rates for varying levels of credit risk.  For example, its current agency fee – the MIP – on its core product of the 30-year fixed-rate, single-family purchase loan21 is almost entirely “flat”:  an upfront 1.75 percent plus a per annum amount as shown immediately below:

  • For up to 90% LTV:                                               0.50% for the first 11 years
  • From above 90% to under 95% LTV:                 0.50% for the life of loan22
  • Above 95% LTV:                                                     0.55% for the life of loan

This is truly minimal adjusting for credit risk.  I note that this applies only to loans of “normal” size (i.e., no higher than an amount known as the conforming loan limit, which is adjusted each year to match house price inflation; it is currently $726,200). For “high balance” loans, available only in defined high-cost geographies, an extra 0.20 percent is added to the per annum rate to reflect not credit risk but an apparent policy decision that such larger loans are not as socially worthy. 

Cross-subsidies.  This near-flat MIP means, economically, that the more creditworthy loans are subsidizing the less creditworthy ones.  As described above, the intended target borrowers for the FHA have evolved to where they are a relatively narrow range of riskier borrowers who need the extra help to qualify.  For example, three-quarters of them have a credit score between 620 and 720; nevertheless, about one-sixth have much better credit scores (between 720 and 850).  Members of the housing policy community who support the FHA’s flat pricing believe such lower creditworthiness is a good proxy for families with lower incomes who should receive the covert subsidy contained within the near-flat MIP.  While a general correlation between income and mortgage loan creditworthiness might have been the best way to target subsidies decades ago, it does not come anywhere near close to targeting those lower-income families as well as other alternatives that are technologically possible today (as has already been implemented at the two GSEs, described below).  Here are two examples where the subsidies could and should be better targeted:

  • A riskier loan is originated from a family that has higher-than-average income among FHA borrowers but has managed its finances poorly (e.g., little savings, high credit card debt, bills often paid late), thus receiving a hidden cross-subsidy in its interest rate.  A loan with less risk coming from a family with lower-than-average income among FHA borrowers which has commendably managed its finances conservatively (e.g., high savings, little credit card debt, bills paid on time) then ends up providing a cross-subsidy.  Instead, today, the FHA could measure household income directly, even in comparison to area median income (AMI) – a well-understood technique in the mortgage industry – and then do a cross-subsidy to the lower-income borrowers from the higher-income ones directly, without relying on only a rough correlation to loan creditworthiness
  • Lower loan creditworthiness, being subsidized, can also reflect a borrowing family stretching to buy a bigger house requiring a large mortgage (even to more than $1 million in defined high-cost areas) with the subsidy provided by a family commendably buying a more modest home (relative to its income) that it can comfortably afford and thus generating a mortgage loan of higher creditworthiness.  Today, the FHA could instead vary the MIP to reflect the price of the house being purchased, even versus the area’s median as well, and perhaps do a cross-subsidy to the lower-priced home purchasers from the higher-priced ones. (The 0.20 percent extra interest rate for high-balance loans is a limited step in this direction.

In short, the FHA could do a better job of targeting its subsidies than it does today: it can directly aim at families with lower incomes or a lower house purchase price, all related to area medians; it can also target first-time homebuyers versus repeat ones; and it can even target counties or ZIP Codes that are particularly low-income for help.  Of course, it would also need to target who provides all these subsidies. 

Thus, it is my view that the FHA would better execute on its mission if it routinely risk-adjusted the MIP on all its loans in a credible23 manner and then added specifically targeted cross-subsidies as decided by policymakers, all while being transparent about how it works and the results.

The VA – Ditto

The VA operates very similarly to the FHA:

  • It does only a minimal amount of risk-adjusting.
  • This then creates significant cross-subsidies, especially as the creditworthiness range of their borrowers is believed to be broader than that found at the FHA.
  • It already engages in various targeted policy subsidies related to its special mission. For example, the entire agency fee (the funding fee) is waived for borrowers with a “service-connected disability” and certain similar situations. 

Just as for the FHA, the VA could do a much better job of targeting its cross-subsidies, rather than utilizing mortgage loan creditworthiness as a crude proxy to do so, and being transparent about them.

The Two GSEs – A Modernized Risk Adjustment and Cross-subsidy Regime; More Transparency Is Needed

The two GSEs are, of course, hybrids of a government agency (characterized by Congressionally-given obligations, advantages, and limitations24) and a private sector, shareholder-owned corporation (with the usual requirement to have capital and to earn a return on it).   As the two GSEs together provide over half of the $13 trillion in mortgages and are also the focus of the recent controversy over changes to the G-fee grids, I will review their situation in greater detail.   

Risk-based G-fees.  Just as any large private sector financial institution does in setting loan interest rates,25 the two GSEs have their G-fees routinely adjusted for risk using the 81-cell grid previously discussed. Used by more than a thousand primary lenders, this grid is of adequate accuracy while being practical for those lenders to implement.  The new risk adjustments just announced were set to be consistent with the FHFA’s regulatory capital requirement formulae (known as the Enterprise Risk Capital Framework, or ERCF), which were themselves set in 2020.  Somewhat controversially, that regulatory requirement, established under the previous FHFA director Mark Calabria, has non-economic factors embedded within it as well.  The prior risk-based adjustment to the GSEs’ G-fees had been set about a decade ago based on historic loss statistics at that time, with some revisions implemented in 2014.  The FHFA, in the May 15 RFI, stated that it was this switch that resulted in the risk adjustments being flattened – i.e., being raised for most low-risk mortgages and reduced for most high-risk ones – and that it was especially impacted by certain non-economic factors embedded in the ERCF.  This is in direct contrast to the criticism that the flattening was a specifically intended policy to redistribute wealth or income.

However, there has been little to no transparency to demonstrate that the flattening did in fact result from the switch to using the ERCF to set the specific amount of risk adjustment in each of the 81 cells, or to determine how accurate that risk-adjusting has been in recent years.  Hopefully, the RFI process, via the resulting FHFA response to the comments submitted, will provide transparency on all this.26 

There is one significant exception to this risk-adjusting:  borrowers who qualify to be in one of the GSE programs to support LMI borrowers are subsidized, and therefore others need to pay a rate higher than economically justified to provide the subsidies.  This issue will now be discussed. 

The Cross-subsidy Implication of Congressionally-mandated and Other LMI Programs.  One aspect of the hybrid nature of the GSEs is that their charters say, right up front, that they should earn, for “activities related to low-and moderate-income families” a “reasonable economic return that may be less than the return earned on other activities.”27  The regulators and the two companies have interpreted this over many years to mean that, for such activities, the two companies will not lose money but neither will they necessarily earn a private-sector level of return.  Such low-return LMI programs first show up in two Congressionally-mandated programs.  One is the long-standing Affordable Housing Goals program (from 1992), and the other is the more recent Duty to Serve (DTS) program (from 2008).  Both require, in practical terms, for the GSEs to reduce G-fees below normal risk-adjusted levels in order to reach the volume and program targets28 set by the FHFA but – as per the charter language – still making at least a minimal profit. 

There are additional programs required by the FHFA, as conservator, that also fall into this category.  All are designed, in one way or another, to help homeownership among LMI families, just as specified by the charters. A recent example is when the FHFA announced a reduction in G-fees to “first-time homebuyers at or below 100 percent of area median income (AMI) in most of the United States and below 120 percent of AMI in high-cost areas.”29 

However, Congress did not address the issue of how two shareholder-owned companies, which need to earn a market return in aggregate on their shareholder’s capital, are supposed to make up for the low returns on such LMI programs.  Presumably, they need to earn a higher-than-normal return on other activities as an offset.  The framework to view this, in my experience, is that the under-market returns of those LMI activities represent the uses of a pool of cross-subsidy dollars – specifically how many dollars of subsidy would be needed by those LMI programs to earn a market return on them.  The question then becomes, what products or set of customers need to be providing a higher-than-market return to create the dollars to put into that subsidy pool?

The Sources of Cross-subsidy.  There are two places to look for cross-subsidy sources.  One is GSE activities outside of the single-family guarantee business.  For example, the multifamily business is regarded as a high-return activity, and thus potentially could be used to provide subsidy dollars.30  But such programs are limited, especially as the single-family guarantee business of the GSEs is by far their largest, so the second place to look – high-return activities within the single-family business – must dominate.31 

The FHFA has, over time, been somewhat transparent as to where it sources the majority of the subsidy pool:  from “mission-remote”32 products rather than a typical loan issued to the usual GSE borrower.  These are single-family loan products considered to be outside of the GSEs’ core mission – which is centered on helping a middle-class family buy a first home or afford a middle-class one.  They are: (1) second home mortgages, (2) investment property mortgages, (3) cash-out refinance loans, and also (4) high-balance loans (i.e., those with a principal amount over $726,200 in 2023 33).  These loans, without special treatment to increase their G-fees, would benefit from the low funding cost enjoyed by the GSEs due to their receiving government support to back their guarantees.  However, it is very hard to argue from a public policy viewpoint that taxpayers should indeed be implicitly subsidizing those types of loans.34  Thus, the FHFA over the years, including recently, has intentionally increased the G-fees on these mission-remote products (in effect, this takes away some or all of the benefit of government support from the borrowers of those mission-remote products), which then collectively generate a cross-subsidy pool

It would be good to have full transparency, where there is little today, about how all the cross-subsidies work.  It should be an FHFA objective to collect the necessary data to do such reporting. Politically, for example, such transparency could do a lot to allay the recent criticisms that the average family borrower unfairly provides the subsidies, rather than mission-remote products. 

Credit Quality of LMI-centric Program Loans.  The recent criticism is also that the G-fee changes which took effect on May 1 were designed – according to its critics – to cross-subsidize high-risk borrowers by overcharging low-risk ones, and thereby will create more high-risk and underpriced loans that will inevitably generate large losses, including in the various LMI-centric programs.

The first problem with this accusation is that we lack the data to assess the riskiness of typical LMI program borrowers.  As an example, with respect to one of the most recent such programs – reducing G-fees for first-time homebuyers with below-area median incomes – the borrower’s credit risk is not even considered in terms of determining who participates in it. 

In fact, the reality is that today the GSEs, with the support of the FHFA, have a range of acceptable credits far different than before 2008.  They no longer purchase mortgage loans with risky product features (e.g., teaser rates, low/no documentation, 40-year maturities, etc., consistent with their adopting certain Dodd-Frank Acts’ product restrictions that apply to private sector lenders), and they no longer try to compete with the very loose credit of the private label securities market that existed back then.  It seems to me that the many reforms to the mortgage system made post-2008 via Dodd-Frank have really been quite effective in avoiding such practices since.  Thus, LMI borrowers still need to meet appropriate credit criteria to qualify for GSE funding; if they do not meet the required GSE credit criteria, then the FHA is there to potentially lend to them, given the latter’s higher-risk focus. 

The combination of LMI-centric program qualification criteria and today’s strong GSE credit underwriting criteria results in focused programs that, as a general rule, have high LTV ratios (i.e., over 90 percent, sometimes even over 95 percent) – which absolutely pushes the credit risk up; but then borrowers under those programs are also required to have high enough credit scores to push the risk back down to where it is a GSE-acceptable credit risk.  In fact, a recent FHFA publication35 on major affordable goals programs shows that, excluding lower-risk refinancing products, this is exactly what is happening:  the average LTV ranges from 90 to 94 percent, which are risky levels, but then the average credit score ranges from 735 to 743, which are surprisingly conservative levels (735 is considered the high end of “good” and 740 is considered the bottom of the “excellent” credit score range36). 

Thus, the mortgage loan to the average LMI program borrower is nowhere near anyone’s definition of unduly risky or akin to the poorly underwritten loans issued before 2008.  To sum up, the GSEs’ LMI programs, which receive cross-subsidies to enable lower G-fees, do not specify that the beneficiaries have high credit risk, and the operation of those programs – based on the data cited above – doesn’t result in it either.  In the post-2008, Dodd-Frank-compliant mortgage markets, the GSEs really do leave the highest risk lending to the FHA and parts of the private sector, where it belongs by design.

A Long-term Increase in G-fees?  I have written before37 how the ERCF, established in 2020, requires much more capital than the predecessor used by the FHFA in conservatorship (known as the Conservatorship Capital Framework).  That implies that the average G-fee has to rise significantly.  Implementing a small portion of that required increase was behind a similarly small amount of the G-fee pricing grid changes implemented on May 1, although this may not have been well understood by critics.  However, the May 15 RFI has clearly and directly stated that this is in fact what is going on – i.e., the new pricing grids do indeed increase interest rates on the typical borrower; however, this is not to subsidize high-risk borrowers but to increase the returns earned by the GSEs by a portion of what is needed to be consistent with the ERCF’s higher level of capital required.  The RFI will allow commenters to directly address the implications of this.


From the above discussion, it is clear that the FHFA is generally running what seems to be a well-considered regime of setting GSE G-fees. To recap, that regime has the following key features: 

  • G-fees are routinely risk-adjusted consistent with proper large financial institution safety and soundness.  This ensures that revenues are aligned with the risk of the underlying mortgages, a key component of safe and sound operation – and also is respectful of the taxpayers who support the GSEs via the PSPA agreement.
  • GSE LMI-related programs are increasingly targeted, with below-market pricing to subsidize them. This is right in line with the mission specified in their charters (i.e., the legislation that established them).   
  • A pool to fund cross-subsidies is funded by raising the interest rate on mission-remote products.
  • The GSEs will generate a commercial market return on the capital used by its single-family activities.  Since the ERCF calls for so much more capital than its predecessor, the necessary increase in G-fees38 is being implemented only over an extended timeframe to avoid disruption.

But without much greater transparency, it is unclear how much this regime actually operates as the FHFA says it does.  In short, transparency is needed to verify that, in reality, it works as described above.  This kind of sunlight will also stop unfounded accusations of distortions and wealth redistribution. 

The FHA and VA should move in a similar direction.  To that end, they should39:

  • Begin to reasonably risk-adjust all insured mortgages with roughly the same level of accuracy (and ease of implementation by primary market lenders) that the GSEs employ now.  This will eliminate the poorly targeted cross-subsidies that result from almost totally “flat” pricing. 
  • Develop carefully targeted programs to determine who will receive subsidies, as well as who will provide the funding for them by paying higher interest rates than required. 

Lastly, the FHA and VA need to develop quality transparency and reporting to demonstrate how these recommendations are being implemented. 

In conclusion, a proper approach to reasonably risk-adjust all government agency fees –combined with charging carefully targeted borrowers a lower interest rate funded by other carefully targeted borrowers paying a higher interest rate – would be far better public policy than what exists today.  It was not practical to do this decades ago when information processing capabilities were much more limited, but today it is very possible.  Such a modernized regime would treat everyone – including the taxpayers who support all four agencies – fairly while producing a much more accurately targeted set of cross-subsidies determined by policymakers to be good public policy.  Adding quality transparency to the process so the public can see how it all works would complete the picture and eliminate any concerns about hidden agendas.


[1] There is a fifth loan agency related to rural housing (part of the U.S. Department of Agriculture, or USDA).  However, its small size has resulted in it generally being excluded from most housing finance policy discussions.  Another agency, the Government National Mortgage Association (GNMA), or Ginnie Mae, is well-known but not included here because it has a significantly different function. Specifically, it is a non-lending unit of the U.S. Department of Housing and Urban Development (HUD) that operates as a common securitization utility to facilitate the issuance of mortgage-backed securities (MBS) to finance the loans insured by FHA, VA, and USDA. 

[2] As measured by dollars outstanding.  See the Urban Institute’s “Housing Finance at a Glance,” April 2023, Page 6, where the $8.9 billion of “agency MBS” accounts for almost 70 percent of all first-lien (i.e., excluding home equity loans) financing.

[3] For FHA and VA loans, this description is a simplification of how the process works, as their process also involves Ginnie Mae.  However, I will use this simplified view, as the topic of this article is not impacted by the more complex underlying process.

[4] The agency fees, somewhat confusingly, go by different names: a guarantee fee (G-fee) in the case of the two GSEs, the mortgage insurance premium (MIP) for the FHA, and the funding fee for the VA.

[5] An upfront fee translates approximately into a per annum fee by dividing it by four to five, based on an assumption about the expected life of a loan and the level of interest rates at the time.  So, a 1.00 percent upfront fee economically translates into a per annum 0.20 to 0.25 percent fee.  I note that the primary market lender will usually enable the homeowning borrower to pay any upfront agency fee in the form of a higher per annum interest rate. 

[6] All delinquency rate data are from the Urban Institute “Housing Finance at a Glance: Monthly Chartbook,” April 2023, Page 29.

[7] This would be an after-tax return, as the GSEs pay federal income tax like any corporation; they are exempt from state and local corporate income taxes. 

[8] The agreement is known as the Preferred Stock Purchase Agreement (PSPA).  It promises the Treasury will inject equity into each of the two companies so net worth never drops below zero, up to certain limits.  The PSPA is strong enough that investors assume that the guaranteed MBS has nil credit risk.

[9] This validates the comment heard from time to time that the top of the FHA credit box – meaning its best quality mortgages – overlaps with the bottom found at the GSEs.  One view, with which I agree, is that this makes the FHA in its entirety a risk-adjusted pricing vehicle, i.e., borrowers who cannot qualify for the lower cost of a loan obtained through the GSEs end up with the FHA where the cost of the loan is higher.   This is, however, a crude mechanism for such risk-adjusted pricing. 

[10] The FHA is required to try to keep a limited reserve, which is not the same thing as capital.  It is also not required to earn a return on that reserve.

[11]  I have previously estimated that the lack of needing to earn a return on capital is worth about 0.30 percent. (See  The greater support to the guarantee (full faith and credit vs. a legal contract to support net worth) has proven to be worth about at least another 0.10 percent, and Congress levies a special tax on new GSE mortgages – but not on FHA or VA – of another 0.10 percent.  However, this lower cost of funding through the FHA by about 0.50 percent is more than offset by the higher loan losses at the FHA, so the cost of a loan through the FHA is, on average, considerably more expensive than through the GSEs. 

[12] Creditworthiness depends on an educated estimate of the risk of loss.  The history is that the accuracy of those estimates can vary tremendously.  Additionally, in practical terms, consumer lending will often measure risks and set resulting pricing in “buckets,” i.e., by aggregating loans with similar characteristics. It is often simply too expensive to price each individual loan on a custom basis. 

[13] Obviously, the exceptions to this are defined programs with a social dimension involved, often in response to legal or regulatory requirements.

[14] The reason is that such flat pricing would tend to attract excessive amounts of high-risk loans, which are underpriced, and a disproportionately small amount of low-risk loans, which are overpriced.  The resulting high-risk credit portfolio, which would likely be earning inadequate returns for its risk, is exactly what bank regulators would not want to see at any institution they regulate.

[15] While the GSEs use many inputs to estimate the specific riskiness of a particular mortgage loan, the largest, most prominent, and most easily available are the credit score and the LTV ratio. Based on this, the FHFA has long used a grid of just these two variables to adjust the level of G-fees, as a simple, implementable yet adequately-accurate way for primary market lenders to know in advance what it will be so they can price specific loans accurately.  (The FHFA’s initially-announced changes added a third variable to the grid – the debt-to-income-ratio.  However, the primary mortgage lending industry had a very adverse reaction based upon implementation difficulties and high costs versus what it believed was a marginal improvement in estimating credit risk.  The FHFA, in response, first delayed and then, on May 11, withdrew the DTI component.)

[16] This is symbolic legislation, as there is no chance it will be passed by the Democratic-controlled Senate.

[17]  See “Fannie Mae and Freddie Mac’s New Pricing Is Not Punishing Those with Better Credit:  Follow the Numbers,” dated April 27, 2023, by Jim Parrott and Janneke Ratcliff.  Note specifically the second chart, which shows the classic right-side-up G-fee pricing grid still in place. A key point made in the article is to properly include the cost of mortgage insurance, which is exactly right.

[18] This phrasing – “low and moderate income” – comes directly from the two GSE charters, i.e., the legislation that created them.  It is discussed below.

[19] See FHFA’s new release:

[20] See

[21] The FHA offers all sorts of loans that have much less volume – refinance loans, 15-year or shorter fixed-rate loans, adjustable-rate loans, etc.  The core 30-year purchase loan is, however, the most prevalent and also important for policy analysis, which is why I focus on it.

[22] The risk-adjusting for this category of LTV calls for the higher risk to generate extra revenue equal to the 0.50 percent per annum for the years outstanding beyond the eleventh year.  As historically most borrowers do not own a home or keep the same mortgage for longer than 11 years, this extra charge will actually amount to zero in perhaps most situations.  It is an example of how minimal the risk-adjusting truly is. 

[23] A secondary market mortgage agency like the FHA needs to implement risk-adjusting in a manner that is implementable in a practical way by the hundreds of its associated primary market lenders.  The approach used by the FHFA and GSEs – a simple two-factor (i.e., LTV and credit score) pricing grid with 81 cells – is certainly a tested approach that could be adopted.  It might even be reasonable to have fewer than 81 cells given the FHA’s narrower range of borrowers. 

[24] These are found in legislation that is usually referred to as the “charters” of each of the two GSEs.

[25] Interestingly, prior to the Great Financial Crisis of 2007 to 2008, the GSEs did not do risk-adjusting, reflecting more of a government agency approach (i.e., not being concerned with classic private-sector profit maximization in the single-family business). Risk-adjusting began only during the GFC and was mostly implemented after the GSEs entered conservatorship under the control of the FHFA. Not doing risk-based pricing, as it undercharges for high-risk loans, would be expected to lead to an undue portion of such riskier loans, and vice versa for low-risk loans.  The result would be a skewed large, high-risk loan book with inadequate revenue to absorb that risk, something very undesirable to regulators.  The FHFA, based on comments made by its leadership over the years, believes this is one factor that caused the GSEs to fall into conservatorship and, as a safety and soundness regulator, wishes to avoid a repeat. 

[26] One measure of matching pricing and risk would be for the FHFA to show the returns earned on each “cell” of the GSE pricing grid.  If all of them are roughly equal, it would show that the risk adjustments are consistent with the ERCF and do not reflect a redistribution of income or wealth.  A separate analysis to show how accurate the ERFC is in measuring risk would show the difference between the ERCF requirements and those based solely on the latest statistical history of losses. A more sophisticated analysis would also show the dispersion of risks within each cell, rather than just the average, to also demonstrate how accurate the risk-adjusting is when only being based on the two factors of LTV and credit score.

[27] See the first section (301) of the Freddie Mac charter,  The same is true for Fannie Mae. 

[28] The DTS program in particular has very detailed program targets in large numbers.  When first implemented, there were more than 100 specific targets.

[29] See the first listed category in the FHFA announcement dated 10/24/22.  

[30] The multifamily businesses of the two GSEs are, unlike the single-family business, competing more broadly with the private sector, which provides most of the lending in this market.  Thus, GSE multifamily pricing is based mainly on competition with private sector lenders, and the historic result in the last decade is that they earn an above-market return.  Economically, this is evidence that the two GSEs do not pass through to borrowers all the economic value of government support enjoyed by the GSEs in multifamily lending, where they definitely do in single-family.  In fact, since about 2013, the FHFA limits the volume of business that the two GSEs can do in the multifamily business to ensure that the advantages each has from government support do not unfairly allow them to take market share away from the private sector.

[31] Prior to conservatorship, this was not true.  At that time, the discretionary investment portfolios of the two GSEs, which peaked at over $1.5 trillion and which were funded by unsecured borrowings at unusually low, near-Treasury interest rates due to the implied guarantee then enjoyed by the two companies, was the largest source of profits.  This was considered by many to be an abuse of the Congressional design of the GSEs, as it benefitted mainly shareholders and management, rather than mortgage borrowers. This allowed the single-family G-fee to be set by management mainly to maintain strong political support in Congress to guard against, among other things, the large investment portfolio profits being reduced or eliminated by changes in legislation or regulation.  The result was a low and flat i.e., not risk-adjusted, G-fee.  Consequently, the GSEs produced a below-market return on the single-family business, as it was so heavily cross-subsidized by the investment portfolios.  This has all been ended during conservatorship, and so the single-family business overwhelmingly now has to have its own self-sustaining economics.

[32] This is not an official FHFA label for such products, but one that I have found reasonably accurately describes them. 

[33] The GSEs are given a maximum loan amount each year, with exceptions for higher amounts in certain defined “high-cost” parts of the country.  In 2023, $726,200 is the maximum amount applicable generally, and $1,089,300 the maximum in high-cost areas.  Loans between these two amounts are sometimes called “super-conforming” or “conforming high-balance” loans.   

[34] It is a justifiable argument that the GSEs should not purchase such products at all.  However, they currently are permitted, and are being utilized to provide the cross-subsidies needed to support LMI activities.

[35] See FHFA’s 2022 Housing Mission Report, Table 10 on Page 15.

[36] See Credit.Org, “What’s a Good Credit Score?  Credit Score Ranges Explained.”

[37] See “Current GSE Guarantee Fees Are Too Low to Be Consistent with Regulatory Capital:  Does This Mean a Large Increase Is Coming?” March 13, 2023.

[38] This presumes that the capital requirement specified by the ERCF is not revised downward.  I have previously recommended that it be revised down significantly to be consistent with the risk of the two GSEs as revealed in the annual official Dodd-Frank Stress Test results.

[39] Legislation would likely be needed to fully accomplish such a change in approach.

Donald H. Layton

Donald H. Layton is a Senior Visiting Fellow from Practice. Prior to joining the NYU Furman Center, he served as a Senior Industry Fellow at Harvard’s Joint Center for Housing Studies, where he wrote extensively about the Government Sponsored Enterprises (GSE) of Freddie Mac and Fannie Mae and more broadly on housing finance. Before his stint in academia, Layton was the CEO of Freddie Mac from May 2012 until June 2019, where he championed the development of Credit Risk Transfers, one of the most significant reforms to the housing finance system in decades.

« Previous | The Stoop | Next »