By the Numbers
Loss coverage and OAS point to pockets of value in CRT
Chris Helwig | February 5, 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.
Investors in Fannie Mae and Freddie Mac Credit Risk Transfer bonds have seen fast prepayments in recent months. And while these speeds often have left pool delinquency rates elevated as a percentage of remaining principal—locking bonds out from principal payments—they have deleveraged the structures. As credit enhancement builds, loss coverage on bonds rises. And a combination of loss coverage multiples and OAS can provide a useful benchmark in measuring risk and reward in CRT. Seasoned mezzanine classes of CAS and STACR deals stand out.
In particular:
- Seasoned STACR high LTV M3s may offer some of the most attractive relative value given relatively wide nominal spreads, attractive loss coverage multiples and thicker slices of risk
- 2016 vintage CAS high LTV M2s also offer attractive loss coverage multiples and wide spreads. Given the premiums they trade at, extended lockouts associated with delinquency triggers should contribute to further IO extension
- Further down the stack, post-COVID, Freddie Mac STACR B2s appear to offer the highest loss coverage multiples at the widest spreads. Higher initial credit enhancement and measures taken by Freddie Mac to insulate these pools from forbearance related delinquencies contribute to low forecasted pool losses and higher loss coverage multiples
While CRTs structures are largely homogeneous at issuance, bonds in these deals can exhibit different risk profiles over time as prepayments de-lever the capital structure. One metric that can be used to evaluate the credit risk embedded in a CRT bond, or any levered mortgage credit instrument, for that matter, is the ratio of a bonds’ credit enhancement to the expected loss on the pool of collateral, commonly referred to as a loss coverage multiple. For instance, if a bond’s credit support is only sufficient to withstand exactly the amount of expected loss for the collateral, it would have a loss coverage multiple of 1.0. All else equal, bonds with similar loss coverage multiples should have roughly the same amount of credit risk. However, the multiple is a measurement of the bond’s exposure to its first dollar of loss. Typically as the multiple rises, so does the rating on the structured debt.
Unsurprisingly, when mapping loss multiples against OAS, bonds with higher loss coverage multiples generally trade to tighter spreads while those with lower multiples to wider ones. (Exhibit 1)
Exhibit 1: Mapping risk and return across CRT
Source: Intex, Polypaths, Amherst Pierpont Securities
While the broad trend is that spread increases with risk, there is meaningful spread dispersion across bonds with similar amounts of loss coverage. Admittedly, ratings, structural leverage and convexity will all factor into the spread at which bonds with similar multiples will be priced and investors need to weigh the combined effect of all these metrics when mining for relative value across CRT. Additionally, certain seasoned B classes are true first loss securities where this framework is not applicable. With that said, there are profiles that offer comparable loss coverage multiples and tranche thickness that trade at wider spreads than comparable bonds.
Based on this methodology it appears that one area of the CRT market that may offer attractive relative value are seasoned last-cash-flow bonds. Particularly M3 tranches of seasoned Freddie Mac STACR trusts backed by loans with original LTVs greater than 80 and M2 classes of Fannie Mae CAS deals backed by comparable loans issued in 2016. Fast prepayments coupled with strong borrower and housing fundamentals have had the combined effect of deleveraging these deals and reducing expected losses. As a result, loss coverage multiples have grown massively in certain classes of CRT in recent months. For example, STACR 2015-HQA1 M3 has a forecasted, pool level, cumulate loss of 14 bp and 3.9% credit enhancement, producing a loss coverage multiple of 28.9 times expected losses. The tranche is 4.9% thick so not only does it have significant credit enhancement to insulate investors from the first dollar of loss but is also de-levered and would take many more multiples of expected losses before the tranche would be written down entirely. Based on forecasted pool level prepayments and losses at a dollar price of $103.5, the bond offers 304 bp of nominal spread and a 298 OAS. Embedded in this assumption is the fact that prepayment speeds will slow materially on the pool as the forecasted 1-year speed on the pool is 28 CPR compared to a January print of 46 CPR. However, the deal is currently failing its delinquency trigger test, which will drive the cash flow to extend regardless of a slowdown in speeds. That benefits the IO embedded in the bond. Other bonds with similar profiles that appear to offer attractive relative value are 2016 vintage Group 2 M2 classes of 2016 vintage Fannie Mae CAS deals. These bonds all have estimated loss coverage multiples of greater than 25 times expected group losses, are between 270 and 300 bp thick and trade to nominal spreads between 290 and 375 bp with OAS ranging from 284 for 367 bp based on model assumptions.
Further down the capital structure, STACR B2 classes backed by high LTV collateral appear to offer the highest loss coverage multiples at the widest spreads. The reference pools backing STACR deals issued in the latter half of 2020 were scrubbed of any loans where the borrower was either delinquent or was marked by being in a payment forbearance plan by the servicer. Additionally, credit overlays implemented in the wake of the pandemic likely drove tighter underwriting of many of the loans backing these trusts. This coupled with an overall constructive view on home price appreciation along with the presence of mortgage insurance on the loans leads to low expected pool losses on these deals. Additionally, in response to the pandemic, Freddie Mac increased initial credit support to B2 classes. The combined effect of all these factors is that these bonds have estimated loss coverage more than four times expected pool losses with forecasted nominal spreads of between 726 and 754 bp. As in the seasoned bonds, these spreads are reflective of a forecasted slowdown in prepayment rates although the disparity between model and actual speeds is far less pronounced in this newer collateral than in more seasoned trusts.