Goldhaber Policy Services, LLC

Developing Strategies & Shaping Policy for Effective Outcomes

Fannie Mae and Freddie Mac Guarantee Fees Comment Letter

August 26, 2014
Mr. Joseph Prendergast
Manager of Policy Research
Federal Housing Finance Agency
Office of Policy Analysis and Research
400 7th Street SW, Ninth Floor
Washington, DC 20024

Re: Fannie Mae and Freddie Mac Guarantee Fees: Request for Input


We are pleased to provide input and share perspectives in response to Fannie Mae and Freddie Mac Guarantee Fees: Request for Input. 1

As authors of the letter that follows, Mike Molesky brings many years of risk analytics and modeling and Mark Goldhaber has years of housing and mortgage policy expertise. The ongoing willingness of FHFA to ask both technical and policy questions should ultimately lead to better answers on appropriate  methodology and level of guarantee fees; for this FHFA should be commended.

Fannie Mae and Freddie Mac were placed in conservatorship on September 7, 2008 almost 6 years ago; there is no legislative consensus on which path  should be followed on GSE reform, which makes these and other regulatory determinations currently out for comment even more critical in determining the cost and availability of conventional financing to a broad spectrum of potential homeowners.

According to the recently released State of the Nation’s Housing 2014, The average guarantee fee charged by Fannie Mae and Freddie Mac jumped from 22  basis points in 2009 to 38 basis points in 2012. In 2008, the GSEs also introduced loan level price adjustments (LLPAs) or additional upfront fees paid by   lenders based on loan-to-value (LTV) ratios, credit scores, and other risk factors. LLPAs total up to 3.25 percent of the loan value for riskier borrowers and are  paid for through higher interest rates on their loans. 2

On December 9, 2013 FHFA announced revisions to the GSE fee structure consisting of the elimination of the stressed market adder fee of 25 bps, but also  increasing the upfront delivery fees for most loans with additional risk characteristics charged to lenders. These fees inevitably get passed through to the  ultimate homebuyer in increased mortgage costs. Director Watt, as one of his first actions as head of FHFA, put the changes on hold and led to the current  request for input.

In FHFA’s 2014 strategic plan for the conservatorship of Fannie Mae and Freddie Mac, FHFA drawing on their statutory authority under HERA, struck the appropriate balance for managing the Enterprises to ensure that —

  • Each regulated entity operates in a safe and sound manner, including maintenance of adequate capital and internal controls;
  • The operations and activities of each regulated entity foster liquid, efficient, competitive, and resilient national housing finance markets  (including activities relating to mortgages on housing for low- and moderate-income families involving a reasonable economic return that may  be less than the return earned on other activities).

The same balanced approach seems appropriate in restructuring an approach to determining the guarantee fees that are appropriate. This letter will  distinguish the different and significantly better performance of fixed rate purchase loans versus refinance loans and recommended reflecting this difference in  pricing. The result should mean significant reductions in g-fees for those consumers who are purchasing their first home or a subsequent home. The GSE’s  provide critical liquidity for the entire market, but from a policy perspective facilitating first-time homeownership and strengthening the overall purchase market  should be the priority. A stronger purchase market would have positive impacts on the overall economy.

The lending construct in today’s market reflects a return to traditional underwriting. The passage of Dodd-Frank and the implementation of the qualified  mortgage rule fundamentally define a less volatile and lower risk mortgage product, which the Enterprises are now purchasing. Indeed, most of the loans  insured by the Enterprises since 2011 have been in the form of qualified mortgages. Based on the information and data that was made available in the FHFA  requests; it is not clear that the proposed guarantee fees and LLPA incorporate all of the new underwriting dictates. If FHFA wants to strike the correct  balance it is important that the pricing and capital reflect the products with lower volatility.

One of the key advantages stemming from the existing structure of both Fannie Mae and Freddie Mac is their unique ability to pool large amounts of  mortgages across all geographic locations and over time, providing the potential for a superior reduction in loss volatility, and a greater degree of liquidity than  any alternative form of secondary market support, a tremendous benefit to the nation’s mortgage finance system. This benefit, in part, is derived from the fact  that there are only two GSEs. Prior to 2008, when Fannie Mae and Freddie Mac introduced broad based utilization of LLPA’s, the GSE’s had utilized pooling to effectively level the guarantee fees charged. While it may not be possible or appropriate to entirely wind back the clock to before 2008, there is no reason  that the benefits of pooling should not be recognized, along with other factors, to reduce the guarantee fees charged to lenders and passed through to consumers. Without some policy balance, the practical reality is that lenders and consumers in many areas of the country will not have access to a Fannie Mae and Freddie Mac execution. The only avenue will be a FHA/GNMA execution. Low and moderate-income consumers and those neighborhoods should  have more than FHA as an option. The Enterprises, with the benefits that they still enjoy, should not build a guarantee fee framework that is based on return maximization and an overly conservative capital framework. When the G-Fee changes were announced on December 9 and then stopped, the underlying  development was initiated at a time when the prevailing attitude was to increase the G-Fee in order to crowd-in private capital and shrink the government footprint.

While shrinking the government footprint is an important policy objective, maintaining higher than necessary G-fees and LLPA’s will not alone restore the private label market. It will require the creation of an effective private label securitization infrastructure that investors can once again have confidence in. Standardization of reps and warrants, repurchase enforcement, loan data transparency and accuracy are some of the issues that must be resolved to restore investor confidence. In addition, policymakers will need to reduce the loan limits for FHA and the GSE’s if a meaningful private label market is to develop. The loan limits should be reduced in an orderly manner over the next several years.

There is an old saying, “in asking what time is it, I need to know the time, not how the watch is built ” In the case of guaranteed fees and capital assumptions, in fact you do need to understand how the watch is built to assure that you get to the appropriate policy objectives and guarantee fee. For  example:

  • Why aren’t the significant differences in loss rates recognized between fixed rate owner-occupied purchase loans and fixed rate owner-occupied rate-andterm refinanced loans in determining G-fees and LLPA’s?
  • Is the targeted ROE justifiable relative to the nature of the net risk exposure and purpose of the enterprise?
  • With regards the suggested levels of capital used for pricing purposes, is the calculation done assuming no cross-subsidization between  origination years, i.e., is each transaction intended to “stand on its own”? Or is the calculation of capital developed as required for a portfolio exposure where risk is generally spread across years?

The answers to the questions that follow reflect our best insights and experience based on the information and data that was made available in the FHFA  requests for comment.

1.  Are there factors other than those described in section III – expected losses, unexpected losses, and general and administrative expenses  that FHFA and the Enterprises should consider in setting g-fees? What goals should FHFA further in setting g-fees?

Clearly the large pricing differences in the various risk buckets have caught the attention of many analysts. One factor that may have a very beneficial effect in smoothing out some of these pricing differences could be the method used in assigning prepayment speeds. Rather than a general prepayment speed  applied to all loans of a specific time period and location, specific prepayment speeds could be estimated for the specific risk buckets. Based on the authors’ long experience with high risk mortgages, we have found that the lower risk mortgages tend to prepay at higher speeds than high risk mortgages under most  economic circumstances. Moreover, higher FICO score borrowers also prepay faster than their lower FICO counterparts of a given LTV grouping. That being  said, while higher risk borrowers generate higher losses, they also generate longer streams of premium income. Consequently, closer examination of actual  prepayment speeds by risk class over long periods of time should be useful in possibly raising some of the low risk fees and lowering some of the high risk  fees.

A primary goal of FHFA in setting g-fees should be that the fees reflect all of the unique benefits the GSE structure brings to the table through diversification  across all markets and across time. Pricing each transaction as a “stand alone situation” fails to reflect the potential capital efficiencies presented by the  GSE structure.

2.  Risk to the Enterprises increases if the proportion of higher-risk loans increases relative to the proportion of lower-risk loans. This change in mix can occur if lower-risk loans are retained on bank balance sheets instead of being sold to the Enterprises. If more higher–risk loans are sold to the Enterprises, or if the overall mix of originated loans changes, what alternatives, other than risk-based pricing, should be considered?

The attractiveness of GSE MBS has been twofold: 1) holding mortgages in the form of guaranteed MBS provides an excellent method of risk diversification not easily achieved by small or even regional banks; 2) holding mortgages in the form of GSE MBS also requires lower capital. Both advantages however, can be greatly reduced if the costs of packaging and insuring such loans exceed the perceived benefits. In fact portfolio lending has increased in 2013, primarily in larger institutions, and is likely to continue to expand over the near term. The increase in portfolio lending is due in part to the lack of an effective and strong private label security market, which must be reconstituted. With the ongoing expansion of portfolio lending FHFA should in coordination with the Enterprises put tools in place to monitor that lenders are not cherry picking the best loans for their portfolio and subsequently adversely impacting the GSE’s. This is an area where FHFA should coordinate with bank regulators throughout the system.

3.  Currently, target return on capital and the amount of capital largely determine required g-fees. What factors should FHFA and the Enterprises consider in setting target return on capital and amount of capital required? How should the Enterprises allocate capital across risk buckets? 3

Setting both the target for return on capital as well as setting the level of capital required by risk bucket are essential to the successful achievement of the GSE primary goal which is to support a viable liquid, efficient mortgage market in all stages of economic cycles. The GSEs originally benefitted the development of the conventional mortgage market by establishing and maintaining prudent underwriting standards, and giving all markets access to national financing rates. Their Charters give them special privileges, and their national presence gave them both superior geographic diversification and diversification over time to reduce the volatility of their mortgage risk exposure. In this respect, the GSE structure should provide for the most efficient use of capital, and competitive pricing in supporting a viable competitive secondary market.

We have three main areas of concern regarding the proposed GSE pricing structure: 1) the combination of fixed rate owner occupied purchase and rate and term refinanced loans as the so-called “base loan” pricing which will lead to the overpricing of purchase loan risk; 2) perhaps the most important concern, given the implied levels of stress losses driving the large levels of capital, the apparent approach of using single origination year loss estimates to set stress losses and capital for pricing purposes, rather than a portfolio approach (or ten or more concurrent origination years) for the purposes of pricing an ongoing guarantee product; and 3) the choice of targeted ROE, given the significant advantages the GSEs have in terms of geographic diversification, and public purpose. All of these concerns relate directly to FHFA’s stated goals for safety and soundness, and a competitive liquid secondary mortgage market.

To start, it is helpful to first review the relationship between pricing and expected losses by category in terms of the base loan mortgage credit risk. In Table 1 below, we have estimated GSE long-run average losses for 30 year fixed rate owner-occupied purchase loans, using recently released data from FHFA, and other sources, including earlier studies of conventional loan performance. We start with this class of owner-occupied purchase loans because such purchase loans are the key to a healthy housing market. As such, these loans should be the base upon which all other riskier factors should be judged and priced for.

The loss estimates for insured high LTV loans are net of assumed standard MI coverage rates. For ease of illustration, we make the comparisons on loans  with specific average FICO scores. Loans with less than or equal 80 LTV had average FICO scores of 745, while high LTV loans averaged between 710 and 725. These average scores are based on the average of FICOs for loans originated between 1999 and 2010, as given by released GSE data. In addition to the simplifying assumptions regarding FICO, we have also taken the position that MI benefits from private insurers meeting finalized eligibility requirements would not be haircut by the 20- 25% adjustment alluded to in the FHFA analytics paper. Had we assumed such haircuts on MI benefits in our analysis as well, the losses to the GSEs after benefit of MI reflected would have been larger than we show in Table 1.

Table-1First of all, the effect of the proposed changes (see footnote for discussion on estimation of fees used in Table 1) is to actually lower the amount of fees charged on owner-occupied purchase loans for all of the categories selected above. However beneficial this would appear, the pricing remains unusually high relative to the risk involved. One can see that for low LTV loans the LLPA accounts for more than 80% of expected loss. For high LTV, the LLPA covers expected loss plus a lot more. When the annual G-fees are added to the mix, the ratios of total fees relative to expected losses go substantially higher. Under the proposed changes, compared to 80 LTV loans, the ratio of G-fees earned relative to long-run losses is higher for all LTV categories except 95.01 to 97 LTV. Also note that expected losses for 80, 85, and 90 LTV loans are fairly equal, yet the total G-fees required for the insured loans is substantially higher. The last column of the table reflects the proposed levels of total G-fees relative to the expected purchase loan losses. It is important to note that, except for loans greater than 95, all of the other LTV groups are either just under 3x or higher.

Based on the attached delivery fee schedules for both Fannie Mae and Freddie Mac, the base rate and term refinanced loans. In the FHFA request for comment regarding eligibility rules for private mortgage insurers, there is a similar absence of differentiation between such loans. Ostensibly, both the pricing grid and the MI stress loss assumptions consider the two types of financing to have the same degree of risk, all other factors considered. However,  examination of cumulative default rates based on the released GSE data suggest otherwise. Looking at the performance of these two types of loans by CLTV, FICO and year of origination reveals a substantial difference in default performance. In Table 2, we show the results of the ratios of cumulative default by the above mentioned cuts weighted by the numbers of rate and term refinanced loans in each bucket for each origination year between 1999 and 2010. The differences in performance are not trivial.


Examination of the ratios reveals that the relative differences between rate and term refis and purchase loans worsen not only as one moves to higher LTV groupings, but the performance worsens also as one moves up the FICO scale. Also note the relatively small differences between rate and term refinanced loans and purchase loans for the greater than 95 LTV loans. Typically, most refinanced loans are generally found in LTV groupings lower than the original LTV. Consequently, it is more surprising that ratios for greater than 95 refinanced loans are as high as they are. Historically, there have been relatively few rate and term refinanced loans with LTVs greater than 95.

In addition, not only are there significant differences in default frequency, there are also differences in loss severity for the same LTV level. Several years ago, Robert Van Order while he was Chief Economist for Freddie Mac pointed out that rate-and-term refinanced mortgages were riskier than purchase loans. Generally speaking, refinanced mortgages suffer from inappropriate LTV measurement in that the value of the property is no more than the opinion of the appraiser, without the benefit of a true sales transaction. As a result, the recovery values relative to stated value at origination on refinanced properties that go to foreclosure often average seven to eight points lower than purchase loans. With average loss severities on most 80 LTV loans running between 25-35%, a 7 to 8 point difference in the recovery value can mean about a 30% difference in loss severity. In Table 3, we show the effects of a 7 point difference in recovery values on loss severity by LTV.

Table-3If we then cross-multiply the ratios of default difference (Column 2) taken from Table 3 times the ratios of the loss severity differences (Column 1) we get overall loss factors that are basically 2X and higher depending on LTV. Using fixed rate owner-occupied purchase loans as a base for GSE pricing rather than all fixed rate owner-occupied loans would greatly benefit home purchasers. Without the change, first-time home buyers are paying a higher G-fee than should be required, in effect, subsidizing the cost of GSE insurance for people who already have a home. If the GSEs are to be effective in their primary reason for being, that of ensuring a healthy housing market at all stages of the housing cycle, they need to set their focus on purchase loan financing as the base loan optic.

We now know that expected losses on owner-occupied fixed rate purchase loans are lower than on owner occupied rate and term refinanced loans. But in order to determine how much pricing may be affected, we now need to know what the difference is in the stress losses between the two. Unfortunately, FHFA has not yet revealed what its assumptions are for the stress level losses used in setting capital for the various categories of GSE risk. To get to an estimate of purchase loan stress level losses, we first have to determine what are the levels of stress losses assumed in the current proposed FHFA base loan pricing, and then estimate the levels of stress losses for purchase only loans. Our approach is to apply what we learned in Table 3 to what FHFA has published as the stress losses for private MI eligibility. In this way we might gain some insight on the stress levels assumed for GSE pricing.

First of all, we need to examine the implied level of stress default frequency. GSE stress losses are net of MI benefits, and MI stress losses reflect the benefits that they pay to its policy holders. In Table 4, we once again will focus on data in the 681-740 FICO range given the 710-725 average FICO of insured high LTV loans. We compare the FHFA MI stress losses to estimated long-run average MI benefits paid and longrun average foreclosure frequencies on fixed rate owner-occupied purchase loans by LTV group. These loss rates were taken from FHFA’s Draft Private Mortgage Insurer Eligibility Requirements. As such, it only contains stress loss levels for high LTV loans. However, this data at least gives us some insight as to the fixed rate loan stress foreclosure frequency levels FHFA may be seeking in its proposed GSE pricing grids for various LTV groups. If we divide the FHFA Stress MI Losses by the effective
levels of standard MI policy coverage, we get an estimate of the implied fixed rate stress frequencies of foreclosure by LTV.

Table-4Again we notice the very large multiples between the stress losses and foreclosure frequencies and the long-run average frequencies and losses for the 80s and 90s. This strongly suggests that the FHFA stress levels may have been influenced by the higher frequencies of loss by rate and term refinanced loans. In the next table we will use the information on performance differences between rate and term versus purchase loans from Table 3, and couple it with information on the share of purchase loans by LTV to estimate what the FHFA stress levels suggest about purchase only loan stress default rates.

Table-5Taking the performance factors we observed in Table 3 and weighting them by the relative shares of both we get an estimate of how much rate and term loans affected the combination of the purchase and refinanced loans together. By dividing our estimate of FHFA implied stress default frequencies for each high LTV group (column 4) with our weighted rate and term performance factors (column 3) we get estimated stress foreclosure frequencies for fixed rate owner-occupied purchase loans. As it turns out, these stress frequencies (except for 85 LTV loans) come very close to estimated foreclosure frequencies for loans originated in 2007. We conclude that FHFA is using single book stress losses based on 2007 originations. There is also no mention of adjusting the private MI exposures as the loans season over time. Therefore, it appears that FHFA intends to assign a single loss rate charge for capital purposes over the life of the privately insured loans. And we assume, therefore, that this approach is also being applied to the GSE pricing calculations. However, the choice of using a single book (origination year) estimation of required capital for pricing purposes is far too conservative. Pricing needs to consider the performance of the whole portfolio, not just a single book. Indeed if the pricing assumes that the whole portfolio will perform at the same level of stress loss as the 2007 books have performed, then one is no longer looking at the probability loss as an occurrence that happens once in 100 chances. 6 More realistically, no  management team would continue to write new guarantee business in the same manner after it became evident that the last two or three books (origination years) displayed unmistakable signs of performing so poorly. Over the last fifty years there have no instances even on a regional basis where older portions of a portfolio have performed as poorly as the newest worst case book.

This phenomenon is not particular to only the US home financing market. Indeed, sovereign regulators of private mortgage insurance in Australia, Canada, and Mexico, all have recognized the importance of a portfolio approach in that older seasoned loans exhibit far less volatility in performance than newer loans under stress. Consequently, these sovereign regulators have set capital requirements that lower the capital charge as loans age. This approach is clearly different from the credit support required of a stand-alone MBS security which would not benefit from the cross-subsidization of other concurrent originations as well as older originations. Assuming that there will be several more books of business in the GSE future, the pricing of that risk needs to reflect such cross-subsidization in the form of a portfolio approach to setting capital for pricing.

Based on our experience, the best way to incorporate seasoning effects on capital requirements over the life of the loan, is to examine historical concurrent loan performance of consecutive origination years assembled by FICO/LTV cuts of loans. In the past, we compared single book frequency and loss outcomes to actual outcomes of loan losses generated by portfolios of ten concurrent books over fifteen year periods for combinations of loans, using combinations of six regions or more (out nine US Census regions) to simulate all potential loss possibilities for a nationally distributed portfolio and the prepayment speeds associated with each of those observations. Our past experience with such studies showed that the expected losses on multiple concurrent books are the same as the long-run expected losses on single books. However, as one moves out on the more extreme economic stress situations, the volatility experienced by the multiple book is more limited than that of a single book, because of the better performance of the older portions of the portfolio. Indeed, all of the regional and national housing downturns in the U.S. and around the globe have been preceded by excessive price increases in real estate values, which allowed all of the older loans to build up a larger equity cushion prior to downturn correction. As a result, in all actual high stress cases, the older loans
performed better on a concurrent basis than newer loans which enjoyed minimal if any price appreciation prior to the downturn. In our experience, the worst stress losses of the multi-book observations were less than half that of the single book worst case losses.

In Table 6, we apply what we’ve learned so far regarding the relative differences in expected and stress level losses between 30 year fixed rate owner occupied purchase loans and rate and term refinanced loans, and what we’ve learned about single book and multiple book capital on the capital assigned for pricing purposes. In the top section, we examine what the use of multiple book capital implies for pricing of fixed rate owner-occupied purchase AND rate and term refinanced loans taken together. We calculate the fees in the manner set forth in Figure 2 of the FHFA request for comment, assuming a 9% targeted ROE. The FHFA proposed fee calculations are developed using a single book capital charge based on 2007 loss rates, while the Multi-book fees uses the 2007 10 concurrent book calculated loss rates. The difference in fees range from 26% to 39% lower.

Table-6In the second section of Table 6, we show the results for fixed rate owner occupied purchase loans only. Because the proposed fee schedule combines the risk of both the purchase and rate and term loans, we leave those g-fees the same as in the top section, and then make our comparison to purchase only using multiple book capital. The differences in pricing for owner-occupied purchase loans are substantial with reductions in g-fees required going from 36% to 64%.

The currently proposed FHFA pricing parameters as applied to purchase loans are so over-priced as to cover a whole portfolio of purchase loans under portfolio stress 15 conditions and still have a substantial profit. In Table 7 we apply the FHFA proposed pricing to the actual and estimated conditions of the 1998 through 2007 30 year fixed rate owner occupied purchase loans from 2007 to 2022. The proposed pricing not only would cover all of the targeted stress losses, but could actually easily cover stress losses that were 60% higher and still make a sizeable profit under what would certainly be horrific mortgage credit losses. Given the lower risk of purchase loans and the approach of using a single book approach to setting the capital charges, the FHFA proposed pricing for such purchase loans is extremely exorbitant and damaging.

Table-7Given the importance of purchase financing access to a healthy housing market, making purchase loan risk the primary base for GSE pricing is essential to the well being of our entire economy. With the substantially lower overall risk presented by purchase loans, they should not be lumped in with substantially higher risk loans for pricing purposes. To do so is to have home purchase financing subsidize all other loans with higher risks. Furthermore, requiring substantially more capital than needed for one particular portion of the financial market also has repercussions for the other sectors of the financial markets. Distortions in the allocation of capital make the whole financial market less efficient. With lower more efficiently estimated levels of capital, lower levels of pricing can still afford a sound return on GSE capital. However, the stated ROE targets and the 16 overall calculation of ROE have also given us cause for further concern. In FHFA’s request for comment on GSE pricing, the Figure 2 Illustration of G-fee computation does not offer any inclusion of actual investment return on the assets held as capital. Most financial firms include such income as part of the overall return. Secondly, while FHFA is mulling over ROE targets of either 9% or 15%, other GSEs such as the Federal Home Loan Banks and others earn substantially less, as do life insurance
companies and other financial service companies. Given the huge advantages of the special charters and social purpose of the GSEs, why would such high ROE targets be part of the proposed pricing structure. With more realistic ROE targets, and more efficient capital levels, the appropriate pricing of GSE risk exposure should be substantially lower than is currently proposed. This would be especially so first time purchase borrowers and for lower FICO borrowers in general.

In summary, we would advocate the following to improve GSE pricing of fixed-rate owner-occupied purchase loans:

  1. FHFA should have the GSEs re-run their base loan analysis using only loans that are fixed rate, 30 year, owner-occupied, purchase loans meeting the current QM standard to form the base set of pricing parameters;
  2. utilize prepayment speeds specific to individual risk groups;
  3. set capital levels based on stress losses of multiple concurrent books;
  4. set ROE targets similar to those in the financial insurance sector.
  5. Re-estimate all additional delivery fees for higher risk factors relative to purchase loan credit risk.

Without these adjustments, the currently proposed price adjustments will continue to drive more low income and low down payment borrowers away from  private capital providers and into the federal arms of FHA.

5 & 6. If the Enterprises continue to raise g-fees, will overall loan originations decrease? That is, will Enterprise loans decline without a commensurate increase in private capital? Is it desirable for the Enterprises to charge higher g-fees on low credit score/high LTV loans if it causes these loans to be insured /securitized through FHA/Ginne Mae rather than through the Enterprises?

Further increases in fees could only mean that the assumed stress losses have increased as well, requiring more capital and higher nominal returns. The  higher fees at this moment will shut out more potential buyers from being included in GSE MBS. But if fewer loans are going into GSE MBS, lenders will also get more restrictive on what they should be willing to hold on portfolio as opposed to holding it in GSE MBS. As a result overall originations will decline certainly in the near term.  scores will be able to use conventional financing on a home purchase . This is why building the strong infrastructure that will restore investor confidence in a private label market is so critical.

7. Is it desirable for the Enterprises to (a) charge higher g-fees on high credit score/low LTV loans if it causes these loans to be insured /securitized through PLS or (b) held on depository balance sheets, rather than guaranteed by the Enterprises?

It is not desirable to overcharge large segments of lower risk potential originations. This can lead to cherry-picking by the largest banks, leading to a distortion in the GSE profile, increasing the overall volatility of risk exposure unnecessarily, thereby undermining ability of the GSEs to provide greater liquidity support for the market in times of stress. The benefits of greater diversification of risk are needed by the smaller banks to manage mortgage credit risk exposure. Without such access they may indeed be forced out of the mortgage lending business. Such loans may or may not take place depending on whether larger institutions also serve those same markets. Access to the unrecovered PLS market is not necessarily an accessible option at this time for many small to regional sized banks. Competitive pricing with reasonable loan limits will eventually enable the private label market to emerge.

8. What approaches or alternatives should FHFA consider in balancing increased use of risk-based pricing with the HERA mission requirements of (1) liquid housing markets and (2) acceptability of lower returns on loans made for low-and moderate income housing?

Using multiple book approaches to setting capital and pricing as presented in our answer to above question 3 is one way. Making sure both the expected and stress loss estimates are based only on loans consistent with QM requirements would also be helpful. Estimating specific prepayment speeds for these specific buckets rather than an overall prepayment speed assumption by product would also be helpful. Applying these same techniques to MI capital requirements would also make MI appropriate pricing of deeper coverage requirements on higher risk low LTV loans another way spreading remaining GSE risk more equally between the various risk boxes.

9. Are the ranges of credit score and LTV cells in the proposed credit score/LTV grids used to set upfront delivery fees and loan level pricing adjustments appropriate? Should any of the ranges be broader or narrower, and if so, why?

First of all, there appear to be no differences in the proposed fees for the top two FICO groupings, so we see no reason for the separation into two groups. The remaining credit score groupings represent a fair distinction between FICO score risk groupings. The two key factors to these credit groupings are: 1) the need for differentiation in non-claim termination rates; and 2) the use of multiple book capital assessment for pricing purposes. These same groupings should also apply to the regulations being proposed for private mortgage insurance companies.

12. Are there interactions with the Consumers Financial Protection Bureau’s Qualified Mortgage definition that FHFA should consider in determining g-fee
changes? The Qualified Mortgage definition as set forth by the Consumer Financial Protection Bureau sets out underwriting boundaries that already limit risk and volatility. To be clear, Fannie Mae and Freddie Mac have been given an exemption from following precisely the statutory Qualified Mortgage rule, but where the mortgage loan meets all of aspects of the published rule, that lower risk should be reflected in the guarantee fee charged by the GSEs. (There are product and underwriting elements in the proposed LLPA and delivery-fee grids that do not fit the Qualified Mortgage requirements.) This approach could provide meaningful benefit to the consumer and help Fannie Mae and Freddie Mac serve a broader market segment. This could be an approach to responsibly reduce the upfront delivery fees and allow Fannie Mae and Freddie Mac greater participation in low income low down payment mortgages.

Submitted by:

Michael F. Molesky
Mark Goldhaber


[1] Mark Goldhaber is a principal in Goldhaber Policy Services, LLC, a firm that focuses on government relations and issues management, primarily in the areas of housing, mortgage finance, and financial services. He has previously served as Senior Vice President, Affordable Housing & Industry Relations
for Genworth Financial’s U.S. mortgage insurance business (formerly GE Mortgage Insurance), Vice President of Public Affairs for Federal Home Loan Mortgage Corporation (Freddie Mac), and held various legislative and regulatory positions with the Department of Housing and Urban Development

Michael F. Molesky is a principal in M.F. Molesky & Associates, LLC, which specializes in domestic and international regulatory and capital analysis with regard to mortgage credit risk. Over his career he has worked in several different organizations with different focuses on the mortgage market. He has served as Director of International Regulatory and Capital Analysis with Genworth Mortgage Insurance Corporation, Director of Mortgage Insurance Ratings with Standard & Poors, Vice President for Credit Enhancement at Fannie Mae, Assistant Vice President for Structured Transactions Group with Moody’s, and as a financial economist at the U.S. Department of Housing and Urban Development.

[2] The state of the Nation Housing 2014 p.21

[3] Our answer in response to Question 3, relies heavily on the previous work experience of Mr. Molesky in his various positions with HUD, two major ratings agencies, Fannie Mae, and Genworth Mortgage Insurance Corporation.

[4] GSE Fees for current projections include 20 bps plus 25 bps adverse market fee plus any LLPA as noted for that LTV/FICO bucket. For Proposed revenue projections, GSE fees include only the 20 bps general fee plus the higher LLPAs associated with that LTV/FICO category.

[5] Average loss rates are based on fully documented fixed rate owner-occupied purchase loans.

[6] The correct calculation of having ten consecutive origination years with same level of stress loss
requires one to multiply the probability of each single book times each of the other. So even with a
chance of 1 in a hundred for each book, the probability of having ten consecutive books with the
same high stress loss level is one in 100 hundred million trillion.

Click for Print Version

Comments are closed.