Returns from prepayments, non-QM and forbearance
admin | November 20, 2020
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.
Investors hoping to add something extra to the usual returns in non-agency MBS next year have a few good choices. The market has done little to distinguish one servicer from another, especially in prime jumbo MBS, but differences are there and available for investors. In non-QM MBS, investors should trim exposure to shelves with the highest concentration of loans with low documentation, which may be subject to legal challenge if the borrower goes delinquent. And in legacy MBS, bonds levered to slower prepayments and recoveries of previously written off principal forbearance look set to deliver attractive returns next year.
Prepayment risk in prime collateral make picking the right servicer important
While a steeper yield curve next year may provide some respite from this year’s elevated speeds, prepayment risk will still likely remain paramount for investors in new issue prime securitizations. The market typically does not price based on the servicers impact on prepayment speeds. Investors that allocate to servicers with slow profiles should pick up extra return.
The jumbo market this year experienced a sea change in the convexity of prime loans. Given the same amount of refinancing incentive, loans with higher balances historically have prepaid faster. However, loans in the prime jumbo market with agency-eligible balances (jumbo conforming) now exhibit steeper S-curves than loans with small agency-eligible balances (conforming) and loans with non-conforming balances (jumbo non-conforming) after controlling for any risk-based pricing or SATO (Exhibit 1). That looks set to change, adding negative convexity to jumbo non-conforming balances.
Exhibit 1: Agency-eligible jumbo loans in PLS trusts exhibit the steepest S-curves
One of the primary drivers of faster jumbo conforming speeds in 2020 was the decoupling between conforming and non-conforming rates. In November of last year the spread between conforming and non-conforming rates was roughly 25 bp. That spread widened to 90 bp in March and has since narrowed to just 12 bp currently, suggesting that in-the-money speeds on conforming and non-conforming loans may converge next year fueled by a couple of factors:
- Jumbo originators whose origination operations were either temporarily suspended or downsized in the spring appear to have ramped back up
- Monetary policies implemented by the Federal Reserve in response to the pandemic have created large stores of excess liquidity. This liquidity, in large part, has found its way to commercial bank balance sheets, creating demand for loans and securities
- Increased competition for loans between bank and non-bank originators may put continued pressure on the spread between conforming and non-conforming rates which would in turn drive faster speeds on non-conforming loans going forward.
While it has been evident for some time that certain jumbo originators tend to prepay faster than others, to date there has been little to no apparent price concession for deals backed by faster paying originators or pay-up for those backed by slower ones. As such, investors can pick up prepay protection afforded by certain originators at little to no cost and likely outperform the broader cohort.
After adjusting for any incremental risk-based pricing on loans produced by a handful of large prime jumbo originators, loans originated by First Republic have historically exhibited a much flatter S-curve than those produced by others. And while investors may primarily be focused on the prepay protection afforded by these loans, given the outlook for a steeper yield curve next year, investors may benefit from faster out-of-the money speeds on these loans. (Exhibit 2)
Exhibit 2: First Republic loans have exhibited a much flatter S-curve
While certain jumbo loans offering prepayment protection such as high LTV loans or New York-based borrowers are generally difficult to source in size, investors can likely scale their exposure to First Republic collateral as the banks is a significant collateral contributor to deals issued by Redwood Trust under the SEMT shelf as well as securitizing their own production.
Lower exposure to non-QM MBS subject to ability-to-repay challenges
Investors in non-QM MBS should try to lower exposure to securities vulnerable to legal challenge on loans that might go delinquent, especially loans underwritten with limited documentation and only one or a few bank statements. The largest foreseeable impediment to non-QM lending returning to form is a largely untested ability-to-repay framework which may become subject to challenge next year. Ability-to-repay, or the ATR standard, has been subject to criticism since it was introduced as part of the Qualified Mortgage rules enacted by the CFPB in 2014. The main criticism of the standard is that despite providing eight unique criteria that a lender must evaluate when determining a borrower’s ability to repay, the rule still retains a fair amount of subjectivity and room for interpretation.
Under the QM rules, borrowers are able to challenge whether a lender established their ability to repay. And if the borrower makes a successful challenge, it limits the lender’s ability to foreclose on the borrower. Despite a pronounced improvement in borrower performance, the default rate on existing delinquencies will not be zero. Borrowers facing default, specifically those whose loans were underwritten using some form of limited documentation, may be in position to challenge whether the lender effectively established their ability to repay the loan.
Loans that appear to be particularly susceptible to ATR challenges are limited documentation loans backed by shorter-duration bank statements as certain lenders underwrote loans to as little as one month of bank of a borrower’s bank statements. While these loans by and large maintained significant compensating credit characteristics at origination, which may mitigate large-scale defaults, some may ultimately serve as statutory challenges to the ATR rule.
The implications of this are likely twofold. First it seems plausible that lenders who originated low documentation loans that are now seriously delinquent may offer permanent modifications, potentially substantial ones, in an effort to preempt potential ATR challenges from borrowers. At a minimum, these modifications would adversely affect interest-only and residual classes of non-QM trusts and could generate losses for subordinate bonds if loans receive balance modifications in the form of principal forbearance or forgiveness. Secondly, this risk may re-shape the type of loans that get securitized in non-QM trusts potentially reducing the amount of limited documentation, owner-occupied loans and driving lenders to concentrate originations in loans that fall outside of the ATR framework, specifically commercial purpose investor loans where alternative underwriting such as asset depletion or using a debt service coverage ratio based on rental income could not be tested under the QM and ATR rules.
Prepayments and forbearance recoveries look to drive returns in pre-crisis loans
Bonds levered to slower prepayments and recoveries of previously written off principal forbearance appear poised to deliver attractive returns next year in the re-performing and legacy sectors of the market. As it appear that the pandemic will likely result in more of an extension event than a default and loss event, overall prepayment rates should continue to remain depressed next year while recovery rates on forborne principal should remain elevated. Certain profiles in the RPL and legacy market may be levered to one or both of these phenomena and will likely outperform as a result.
Some of these profiles include:
- Freddie Mac SCRT BXS tranches that are comprised of both interest-only and principal-only components
- Legacy RMBS interest-only bonds backed by pay-option ARM collateral, particularly those where the loans and bonds are backed by different floating rate indices
- Legacy RMBS mezzanine bonds where principal forbearance has been previously written off and the cash flows receive outsized returns from subsequent recoveries of forbearance, including bonds that have been written off but may receive cash flow from forbearance recoveries
- Bonds in RPL trusts that are backed solely by non-performing forborne principal
A thematic and potentially scalable trade for mortgage credit investors next year may be to pair interest-only and principal-only like profiles to gain exposure to slower prepayments and elevated forbearance recoveries. Pairing these profiles intuitively makes some sense as the long IO trade implicitly is a view on a slower economic recovery while the long PO trade takes a constructive view on broad-based US housing fundamentals so the paired positions acts as somewhat of a hedge against one another. And there are likely scenarios where both sides of the trade may yield positive outcomes.
While there are multiple ways to express this trade one exposure pre-packages the interest and principal-only components. BXS classes issued under Freddie Mac’s re-performing securitization channel, the SCRT program, are cash flows comprised of a principal-only class, which is backed in part or whole by securitized principal forbearance and three interest only classes representing interest stripped off the senior bonds, the subordinate bonds as well as excess interest. Unlike the legacy market where by and large principal forbearance is written off and then subsequently recovered, these bonds are actually securitized by the non-performing balance of loans that have received principal modifications. Unlike the legacy market where even a partial recovery is a windfall to bond holders, anything less than a full recovery in these deals would generate a loss.
However, SCRT BXS classes do maintain certain advantages over legacy profiles levered to forbearance recoveries. First, the principal backing these bonds is forborne and not forgiven. Given data and reporting issues rife throughout the legacy market, principal forgiveness which is unrecoverable is often misreported as principal forbearance. Additionally, while holders of legacy RMBS have little to no control over whether the servicer recovers a borrower’s forborne balance, absent a default and subsequent liquidation where the recovery proceeds are not sufficient to pay off the forborne amount, Freddie Mac directs the servicer to recover forborne amounts. And while the forbearance is securitized, the principal-only component is deeply discounted so investors can generate total return through both carry in the interest-only classes and accretion to par as forbearance gets paid back.
Another way investors can get exposure to the slower prepayments are interest-only bonds backed by pre-crisis pay option-ARM loans. Option ARM IO may outperform based on a number of factors. First Option ARM collateral still exhibits elevated delinquency rates relative to other legacy cohorts. Elevated forbearance and delinquency rates should suppress prepayments on option ARM loans, enhancing carry on IOs backed by the collateral. Additionally, option ARM collateral by and large has outsized exposure to judicial foreclosure states like New York, New Jersey and Florida relative to other cohorts. Given the long liquidation lags that already exist in judicial foreclosure states, specifically New York and New Jersey, which will likely only further extend due to Covid-19 related foreclosure moratoriums and court closures, it seems unlikely that these delinquencies will materialize into losses that would cause write-downs and curtail the IO any time in the near future.
In fact, given the amount of built-up borrower equity in the cohort, it seems that there is a greater likelihood of modification than foreclosure and subsequent liquidation on loans that roll into late stage delinquencies. If a greater amount of previously always-performing option ARM loans become modified, historical prepayment rates suggest speeds should slow down materially. Trailing prepayment rates show that always-performing loans have prepaid almost 20 CRR faster than modified re-performing ones, which have paid roughly in-line with non-performing loans where the prepayment was likely the function of liquidation with no loss severity (Exhibit 3). Admittedly, a high concentration of RPLs may mean the coupons on modified loans may be fixed for some period and capped at a certain rate, potentially mitigating the effect of the MTA-LIBOR basis.
Exhibit 3: Option ARM RPLs prepay much slower than always performing loans
Investors can add exposure to elevated levels of principal forbearance recoveries in multiple forms. Legacy mezzanine bonds, zero factor legacy subordinates that have been previously written off and tranches of RPL trusts backed exclusively by principal forbearance an all provide exposure and leverage to elevated recoveries.
Despite the pandemic’s impact on borrower performance, forbearance recoveries continue to flow through to both legacy and re-performing RMBS trusts. Upside to forbearance recoveries still remains a sizable opportunity given nearly $32 billion in outstanding reported forbearance on loans in legacy trusts as well as an estimated $15 billion in potentially unreported forbearance. Over the past six months, certain legacy trusts that have received the largest nominal amounts of forbearance recoveries on average have recovered three quarters of the forborne amounts, modestly higher than long term trends and likely a constructive sign for continued elevate recoveries into next year. (Exhibit 4)
Exhibit 4: Stacking up recent forbearance recoveries by shelf
1 (646) 776-7714