Valuing post-forbearance Ginnie Mae loans
admin | April 30, 2021
This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors.
The Covid-19 pandemic has created a new exposure for MBS investors in loans that go into forbearance. The CARES Act, passed in March 2020, offered forbearance to borrowers simply for attesting to hardship related to pandemic. Most borrowers could suspend payment of principal, interest, taxes and insurance without a mark against them on credit bureau reports. The legislation initially authorized forbearance for up to 12 months, and Fannie Mae, Freddie Mac, and all agencies with loans in Ginnie Mae MBS have extended it to 18 months. Uptake across the mortgage market has been significant, especially in loans backing Ginnie Mae MBS. As these borrowers now start to come out of forbearance, MBS investors face a new set of exposures without a simple precedent for valuation. But there are hints of opportunity.
To navigate the flow of post-forbearance loans over the next few months and beyond, it helps to focus first on Ginnie Mae and set the stage in a few parts:
- Ginnie Mae’s response to the surge of forbearance delinquencies
- Servicers’ choice to either buy out or cure and refinance delinquent loans
- Buyout activity so far
- Activity in new RG pools
- Prepayment behavior in post-forbearance loans
- Differences between RG and post-forbearance custom refi pools
- Potential supply of post-forbearance loans, and
The types of loans likely to dominate the flow, their seasoning, loan balance and deferred payments all suggest post-forbearance loans should be worth more than TBA. How much more is open to reasonable debate. The details follow.
Managing the surge in Ginnie Mae delinquencies
The Covid-19 pandemic over the last year has sharply raised the amount and percentage of delinquent loans backing Ginnie Mae MBS. The majority of the delinquencies reflect forbearance required by the CARES Act. The sector went from $100.9 billion in principal and 5.3% of loans delinquent before the March 2020 start of the pandemic to a peak of $230.7 billion in principal and 12.0% of loans delinquent in June 2020 (Exhibit 1).
Exhibit 1: The start of pandemic in March 2020 brought a surge in delinquent loans backing Ginnie Mae MBS
Ginnie Mae last year recognized the problem these delinquencies might create for investors and the performance of MBS. Servicers can buy loans out of a Ginnie Mae pool if the loan is at least 90 days delinquent. This is a longstanding tool for managing delinquency pipelines. But buyouts create prepayments, hurting performance in MBS trading above par. A surge in delinquencies and potential buyouts would have created significant uncertainty about the timing of Ginnie Mae MBS cash flows and the value of the securities.
For servicers, buyouts can be a boon. Ginnie Mae traditionally has allowed servicers to buy out loans at par and immediately deliver any that cured into a new Ginnie Mae pool. Since low interest rates last spring had pushed almost all MBS to significant premium prices, servicers had the potential to make a lot of money by buying out delinquent loans at par and repooling them into premium MBS. These gains would have come at the expense of investors.
Ginnie Mae last July took two steps to curb the pace of buyouts:
- The agency made it less profitable to buyout and cure delinquent loans by only allowing issuers to sell them through a new non-TBA-eligible pool type, RG
- And Ginnie Mae worked with the Federal Housing Administration to make it clear most borrowers could refinance following forbearance
Buy out or refinance?
The Ginnie Mae action has given servicers two broad paths to follow with loans coming out of forbearance, something likely to happen with increasing frequency as the market passes the first anniversary of Covid:
- Servicers can buyout delinquent loans and, subject to restrictions, sell them into RG pools, or
- Servicers can allow those loans to cure in the pool and later attempt to refinance them through the normal course of business
To understand the risks and rewards of either path for servicers, it helps to understand the basics of an FHA partial claim. While in forbearance, the FHA allows borrowers to suspend payment of principal, interest, taxes and insurance. After forbearance, the FHA allows borrowers to resume regular payments and defer all delinquent payments at no interest. Repayment happens only when the borrower moves, refinances into a non-FHA loan or when the loan matures. Note the borrower can refinance into another FHA loan without repaying the deferred amount, at least if the borrower has left forbearance and made the three most recent payments. The loan consequently can remain in its MBS pool since the loan’s payment terms are unaltered.
The partial claim is the first step in the FHA’s waterfall following Covid forbearance. Most borrowers should get partial claims, since they only need to resume making their old mortgage payments to qualify. The USDA’s Rural Housing Service (RHS) and HUD’s Public Indian Housing (PIH) each offer a program like the FHA’s partial claim, although most RHS loans are likely to be modified since loan modification is higher in the post-forbearance waterfall. The VA is unable to provide funding for a payment deferral program, so most VA loans will likely refinance or be modified.
Selling a buyout loan into an RG pool has specific requirements. The servicer must buy out the delinquent loan, cure it through partial claim or other means, get the borrower to make at least six consecutive payments and can only repool the loan after a minimum of 210 days from the last date the loan was delinquent. The servicer incurs costs for financing, hedging, servicing and any expenses or losses if the loan becomes delinquent again or goes to foreclosure. Return on equity for buying out loans and selling them into RG pools can be substantial but varies widely depending on cost of funds, time to repooling, the price received for the pool and other factors.
Instead of buying out the delinquent loan and repooling it, the originator can try to refinance the loan. First the originator has to cure the loan through a partial claim and keep the loan in its current pool. After three consecutive payments, the servicer can refinance the loan through FHA’s streamline program. Streamline refinancing requires neither a credit score nor an appraisal, and the borrower’s deferred balance transfers to the new loan. The originator would profit from any fees charged for the refinancing and any price premium received from selling the loan in the secondary market. The servicer might not earn as much revenue from refinancing compared to buyout—not every loan will refinance, and those that do will be sold into a pool with a lower coupon than an RG pool and a smaller price premium. But there are also benefits from lower financing and hedging costs and eliminating the risk of owning delinquent loans for an extended period. Many servicers may find the relative simplicity of refinancing attractive.
Servicers of FHA loans, the loans most likely to be bought out, face a complex decision (Exhibit 2). Most delinquent loans are on Covid-19 forbearance plans, which means they will be eligible to refinance three months after curing. Any cured loans that were bought out can only be sold into RG pools. Loans that are not on Covid-19 forbearance are less likely to receive partial claims and cannot refinance for 12 months after curing.
Exhibit 2: What happens to delinquent FHA loans in Ginnie Mae pools
Buyout activity so far
So far, banks have led the way in buying out delinquent loans from Ginnie Mae MBS. Of the $84.4 billion in buyouts since July, banks have taken out $34.3 billion and non-banks $50.1 billion. Banks have bought out 73.4% of their cumulative delinquencies since July, non-banks have bought out 28.4%. The bank appetite almost certainly reflects their lower cost of funds and greater need for earning assets. Non-banks have faced a higher cost of funds and a more limited supply of funding. So far, the potential profits from buying out loans have not compelled non-banks to do it at nearly the rate of banks.
The RG program so far
Ginnie Mae described RG pooling requirements as temporary when announced in July, but they could become permanent. The program would benefit investors anytime delinquencies surge, whether from an economic slowdown or a natural disaster. Leaving the program in place means Ginnie Mae would not need to relaunch the next time delinquencies surge.
The first RG pools came out in February 2021, and through the end of April roughly $3.2 billion of pools have been issued. Wells Fargo and PennyMac constitute the largest issuers with 79% of issuance so far. Other issuers include Carrington, Lakeview, NewRez, Truist, and U.S. Bank. Ninety-six percent of these loans are FHA, while only 2.7% are VA and 0.9% are RHS.
PennyMac and AmeriHome have also sold a larger amount of RG pools on a forward basis. Both lenders retained an option to roll delivery into future settlement dates to mitigate the uncertain timing regarding when loans become RG-eligible. PennyMac’s forward trades have recently sold with roughly 1-point pay-ups over TBA—36/32s for 3.0%s, 37/32s for 3.5%s, 31/32s for 4.0%s, and 26/32s for 4.5%s. Pay-ups have been much stronger for newly issued pools where investors can see and model the actual collateral, although have weakened recently. Pay-ups in these coupons have been bolstered by dollar rolls that are currently flat to negative.
Prepayment behavior in post-forbearance loans
Regardless of whether post-forbearance loans end up in RG pools or get refinanced into new multi-issuer or custom Ginnie Mae pools, the loans may have prepayment features that make them different from loans that never went through forbearance and received a partial claim. In fact, a snapshot of loans bought out of Ginnie Mae pools so far, of loans seriously delinquent and likely to get partial claims in the future and of loans in recent multi-issuer TBA-eligible pools highlights some important differences (Exhibit 3). In particular:
- Most of these loans will likely come through FHA and have a much more stable prepayment profile than VA loans
- Most of these loans will likely come with more seasoning than average TBA loans
- Most of these loans will likely come with smaller loan balances than average TBA loans
- The partial claim will lower the borrower’s effective interest rate
Exhibit 3: Early buyouts have higher FHA share, lower loan size, more seasoning
More stability in FHA than in VA loans
FHA loans have historically refinanced much more slowly than VA loans. This is broadly due to the weaker credit of FHA loans. Even though FHA and VA streamline refinancing involves no credit score or appraisal, FHA loans refinance less aggressively. From March through December 2020, for instance, FHA loans with 50 bp of refinancing incentive and balances above $200,000 only refinanced at roughly half the pace of VA loans. FHA loans with 100 bp of incentive refinanced at roughly 60% of the pace of VA loans.
The prepayment difference between FHA and VA loans was also evident before the pandemic, even after changes were made by Congress and Ginnie Mae intended to slow VA refinancing. Exhibit 4 compares FHA and VA S-curves after the VA refinance reforms but before the onset of the pandemic. FHA prepayment speeds peaked at roughly 50 CPR, while the fastest VA loans reached nearly 90 CPR, which is over three times faster on a monthly basis (calculating the difference in SMM terms).
Exhibit 4: FHA loans prepaid slower than VA loans before the pandemic
The predominance of FHA loans in RG pools and in other post-forbearance refi pools should improve their convexity compared to TBA. The predominance of FHA comes in part because most Ginnie Mae loans are insured by the FHA (55%) and VA (39%), while RHS loans account for only 6% and less than 1% are PIH loans. Second, only the FHA and PIH place their payment deferral programs at the top of the post-Covid waterfall. The VA cannot fund a payment deferral program, so most loans will refinance or be modified. The USDA’s payment deferral program, known as the Mortgage Recovery Advance, is placed below loan modification in the waterfall.
The VA does not offer a payment deferral program because the VA does not have direct lending authority. The VA will permit lenders to offer payment deferrals, but the lender would need to provide the 0% financing. This seems unlikely. Similarly, VA lenders can negotiate a repayment plan with the borrower, but the lender is also financing the delinquent balance at no interest over the course of the plan. It is doubtful many VA borrowers receive these offers. This means most delinquent VA loans will end up with a loan modification or refinance. The VA permits refinancing delinquent loans, recapitalizing the delinquent balance, although the VA must approve each refinance.
The VA share may be slightly larger in the first RG pools since those loans would only have been a few months delinquent when they cured last summer. That makes it possible that some VA borrowers could cure their delinquency with an out-of-pocket payment. But later pools should contain loans that were in deeper delinquency before curing, and fewer of these loans are likely to self-cure.
More seasoning and lower loan balances
Highly seasoned loans tend to refinance less aggressively than newer loans. Ginnie Mae II 30-year 3.5% pools, for example, had more than 100 bp of refinancing incentive in December 2020. The 2019 vintage prepaid that month at 60 CPR while vintages from 2015 or earlier prepaid between 32 CPR and 45 CPR (Exhibit 5). Other coupons show a similar tendency for speeds to tail off with age and burnout.
Exhibit 5: Seasoned, more burned out vintages tend to prepay more slowly
It is usually difficult to source seasoned FHA pools since so many loans get delivered into multi-issuer pools. The RG program should offer unique exposure to seasoned FHA loans. Refinanced loans that go back into multi-issuer pools or that get sold in custom pools should have similar inherent seasoning, but the refinancing process resets the loan age to zero so the seasoning will not be apparent.
Post-forbearance loans also should have a much smaller loan size than recent production, and smaller loans offer refinancing and extension protection. They are also more likely to be first time home buyers, which also tend to refinance more slowly but move more rapidly. However, the added seasoning brings lower current LTVs due to longer amortization and appreciation.
Partial claims may not slow refinancing but may slow turnover
Putting aside its other features, it is unclear whether a post-forbearance loan will refinance FHA-to-FHA more slowly than an otherwise identical loan that was current throughout the pandemic. The FHA permits streamlined refinances after three months, and borrowers do not have to repay any partial claims if the new loan is an FHA streamlined refinance. The CARES Act should have prevented any negative credit reporting on these loans, although credit is not checked during a typical streamlined refinance. It is possible that speeds could be slower since a servicer marketing pools backed by these loans might avoid soliciting refinances that would harm execution of future pool sales, but it could be hard to enforce with third-party originators. Prepayment speeds did seem to slow on loans that were affected by 2017 hurricanes, but it is not clear that experience extends to Covid forbearance. Some investors may assume that loans exiting forbearance will refinance similarly to identical loans that did not go through forbearance and receive a partial claim. That means any premium to TBA must come from other characteristics of these loans.
The presence of the partial claim may have a small effect on FHA-to-conventional refinancing speeds. It could slow FHA-to-conventional refinances for borrowers with current LTVs below 80% and combined LTVs above 80%. Those borrowers can refinance into a conventional loan to avoid paying mortgage insurance but will have to repay the partial claim. That means the combined LTV must fall below 80% and borrowers would forfeit the 0% interest on the deferred amount. The presence of a 0% interest partial claim adds an elbow shift to the loan’s refinance incentive.
The partial claim may increase extension risk for these loans. Borrowers must repay the partial claim if they move, and this could erase a substantial amount of equity or even create negative equity in newer loans with little accumulated amortization or home price appreciation. A partial claim to cure 12-month forbearance on a 4% note rate loan could be more than 7% of the loan’s UPB. This could make it difficult for these borrowers to move, slowing turnover speeds.
FHA post-Covid cures are showing they can refinance
There is evidence that FHA loans that became delinquent during Covid and returned to current status without leaving a pool face refinance frictions but nevertheless can refinance. This comes from a few snapshots of recent FHA experience:
- A snapshot of all FHA loans that became at least 90 days delinquent since March 1, 2020 and their subsequent status
- Prepayment speeds on loans that subsequently cured
- A comparison of speeds on cured FHA loans to speeds on FHA loans that remained currently consistently after March 1, 2020
Loans in Ginnie Mae pools that were current on March 1, 2020 and subsequently became at least 90 days delinquent fall into two categories: loans that never cured, and those that did (Exhibit 6). There is a Ginnie Mae field to indicate the reason a loan exits forbearance, but it is not yet populated. There consequently is no way to separate borrowers that cured with partial claims from those that cured some other way. It is likely most cures were due to partial claims.
Exhibit 6: Over 6.5% of FHA loans that cured have subsequently prepaid
Roughly 6.6% of loans that cured have already prepaid voluntarily, and another 1% subsequently redefaulted and were bought out of the pool. The remaining 92.5% are still in pools—81.5% are current and another 11% have missed payments again but are still in pools. This suggests there is measurable prepayment and re-default risk in loans that receive partial claims.
Of the loans that never cured, 25.6% were subsequently bought out as an EBO and another 7.5% were bought out as mandated by a loan modification. Another 60.5% are still at least 90 days delinquent. Some of these loans were also able to prepay voluntarily. This could indicate a borrower selling the home for enough to pay off the full loan amount including delinquency.
It is also notable that the number of loans that have cured trails the number of EBOs in this population by only 26%. Some servicers have chosen to forego the buyout on a substantial number of loans.
FHA loans that went seriously delinquent and then cured have since voluntarily prepaid (Exhibit 7). Prepayment speeds are calculated based on the number of months of seasoning following the first time the loan cured. These cohorts generally included the largest number of delinquent loans and were chosen to give good representation across coupons and vintages. The exhibit includes only voluntary prepayments speeds and excludes buyouts.
Exhibit 7: FHA loans prepay shortly after curing.
One surprising observation is that the loans can prepay almost immediately, even though the FHA currently requires borrowers to make three payments after exiting forbearance before refinancing. The exhibit assumes the date a loan is marked current coincides with the date the loan exits forbearance, since the delinquency status is the driver of buyout risk. It is possible that the delinquency status is updated on a lag, but that would not explain why the 3-month refi lockout is missing.
By the third month after curing, each cohort was prepaying at least 20 CPR from voluntary prepayments (Exhibit 8). Compared to FHA loans that consistently stayed current after March 1, 2020, speeds in the cured loans are slower. The average prepayment speed for loans that are three months to six months cured is best compared to the average prepayment speed from December 2020 through February 2021 for the comparable FHA always-current cohort. The older vintage cured loans prepaid 30% to 33% slower than the cohort, while the 2019 vintage loans prepaid 36% to 52% slower than cohort. Since the analysis does not filter out differences in loan balance, this factor may be contributing to the slower speeds in the cured FHA loans.
Exhibit 8: FHA cured loans prepaid slower than cohort shortly after curing
These data suggest that loans in RG pools or in custom pools of post-forbearance refi loans could prepay quickly shortly after pooling. A loan that already refinanced and was sold into a custom or multi-issuer pool would have less prepayment risk due to the lower note rate.
Comparing RG and custom refinance pools
Servicers will have a choice of either keeping post-forbearance loans in their current Ginnie Mae pool, buying them out and repooling them into an RG pool or refinancing the loans into a production coupon pool. Only RG pools will be available as distinct exposures in premium coupons, while pools of refinanced loans will come in the production coupon. Apart from the obvious differences in note rate and pool coupon, the loans available to each pool type will be similar. Most should be FHA loans with some level of seasoning, smaller loan sizes and a partial claim attached.
There will be some differences at the pool level itself. The refinanced pool will not reflect the additional seasoning in these loans since the loan age resets to zero at refinance. Those pools will also have longer maturity dates. And the refinanced pools will not be available in premium coupons. A pool of refinanced loans, by definition, would have prepaid quickly following forbearance so may be better suited to investors that expect past forbearance to pose few obstacles to refinancing.
One disclosure advantage for the refinanced loans compared to RG loans is that the partial claim amount should be included in the loan’s combined LTV. The RG pool’s loans will reflect the original and combined LTV at origination. Ginnie Mae added a field to indicate the reason a loan exited forbearance, but it is not yet populated. Even if it were populated it probably will be cleared after a refinance and is unclear whether it would be recorded in an RG pool.
Recent pay-ups for 0% VA pools have been very low, so an investor might find a custom pool to compete on relative value to an RG pool that trades at a pay-up to TBA.
The supply of new pools backed by post-forbearance loans could grow substantially since so many loans became delinquent during the pandemic. This could be RG pools backed by buyouts or new custom pools backed by loans that refinance three months after exiting forbearance. Bank servicers currently have $9.5 billion of FHA loans delinquent 60 days or more, while non-bank servicers have $94.0 billion delinquent FHA loans (Exhibit 9). Only FHA loans are considered since they should make up most loans that are placed in RG pools, and investors looking at custom pools of refinanced loans would stipulate no VA loans in those pools.
The recent extension of forbearance up to 18 months for many borrowers could delay the timeline for these loans to make it into new pools.
Exhibit 9: Potential post-forbearance loan supply
Bank servicers have been more aggressive than non-banks at buying out delinquent loans, but banks may not be large issuers of RG pools. Banks have bought out 71.4% of their total supply of these FHA loans but service fewer Ginnie Mae loans than non-banks, which limits the potential supply. Some bank buyouts will end up with loan modification, since the big banks bought out most loans as soon as they were eligible. It is possible banks may prove more likely to retain these loans on balance sheet or retain any RG pools they create.
There is a larger potential supply of these FHA loans from non-banks. So far, non-banks have only bought out 23.7% of their potential loans, many of which could end up in RG pools. However, the fate of the 76.3% still in pools is uncertain. Some servicers will choose to buy out these loans and put them in RG pools, and others will choose not to buy out these loans and instead attempt to refinance them after three months. And some of these loans will end up receiving loan modifications or liquidate, resulting in a mandatory buyout and subsequent TBA or custom pooling.
The large amount of delinquent loans still in Ginnie Mae pools suggests that many lenders have not been able to secure financing or are choosing not to exercise their buyout option.
The Amherst Pierpont prepayment model can help estimate fair value for RG or custom pools based on collateral characteristics: FHA share, seasoning, average loan size, FICO score, and an assumption about the effect of prior delinquency and forbearance (Exhibit 10). By starting with a TBA pool and then changing its attributes one by one, holding OAS constant, the model can estimate the separate value of each feature. Most loans that refinance are likely to be issued in 2.0% and 2.5% pools, while most RG pools should be issued in 3% and higher coupons.
Exhibit 10: Collateral differences do not fully explain market pay-ups
The theoretical pay-ups are roughly consistent with market levels for pools sold on a forward basis, such as those from PennyMac. But they are well below the levels being paid for issued pools, like those being sold by Wells Fargo. A possible reason for the difference between the model pay-ups and the pay-ups for Wells Fargo’s pools is that investors in those pools are anticipating much slower prepayments due to the past delinquencies of these loans than the assumption in the Amherst Pierpont model.
The Amherst Pierpont model places 5/32s to 13/32s of value for 100% FHA collateral, depending on the coupon. The exhibit also includes two additional estimates of the value of 100% FHA collateral, based on the S-curves presented in Exhibit 4, that result in higher theoretical pay-ups for 100% FHA collateral. These scenarios, labeled “FHA 36” and “FHA 60”. assume that FHA loans refinance slower than VA loans until they reach 36 WALA and 60 WALA, respectively. At higher WALAs FHA and VA loans are projected to prepay identically. Both FHA scenarios assume FHA and VA loans prepay identically when out-of-the-money. For example, the “FHA 36” pay-up on 2.5%s is 19/32s, which assumes the prepayment benefit of FHA loans only lasts for 27 months since the TBA-proxy pool is 9 WALA. The pay-up for 100% FHA in 4.0%s through 5.0%s is generally lower in the “FHA 36” and “FHA 60” scenarios for two reasons—the TBA pool has a lower VA share, and the pools are more seasoned than lower coupon TBA, so the benefit of FHA over VA does not last as long.
The model places a lot of value on the seasoning benefit in lower coupon pools, and then diminishes at higher coupons since the TBA pools are typically more seasoned. The 4% through 5% pools used to proxy TBA are all roughly 30 WALA. Most loans that are placed in 2.0% and 2.5% pools will be refinances, so the seasoning will not be apparent, but may still result in slower prepayments compared to similar coupon pools that were issued in 2020.
Loan size also contributes to the value of this collateral. For example, the Amherst Pierpont model assigns a 9/32s pay-up for loan size in 2.5% and a 10/32s pay-up in 3.5%s. The pay-ups for lower credit scores are positive, although generally smaller than pay-ups for FHA, seasoning, and loan size.
Finally, the Amherst Pierpont model incorporates a view that FHA loans that were previously delinquent and in forbearance will face additional refinance frictions but possibly default at a slightly faster rate. This view is based on prepayment speeds of loans with only one 30-day delinquency that immediately cures and remains current for six months. Prepayment behavior is tracked after six current payments (loans can subsequently go delinquent). These speeds seem consistent with the limited prepayment history of RG pools in March and April.
The model values the effect of past delinquency at 4/32s to 5/32s in most coupons. These pay-ups were not run for 2.0% and 2.5% pools since most loans in those coupons are likely to be refinances, so will have proven that pay history was not a barrier to refinancing. But RG pools, which will typically come in the higher coupons, will only include loans that did not refinance after curing. There is a four-month window during which a loan can refinance before it can be sold in an RG pool. Therefore, these loans may refinance somewhat slower than loans that did refinance.
There is significant uncertainty regarding how these loans might perform, whether in an RG pool or refinanced. Without historical experience to lean on investors in this collateral need to decide whether to accept the risk.