Screening for relative value in CRT last cash flows
admin | February 14, 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.
Credit spread tightening coupled with low rates and fast prepayments have put almost all Fannie Mae and Freddie Mac CRT last cash flow classes above par, and prepayment risk has trumped credit risk when trying to mine for relative value. As a result, higher premium 2016 and 2019 last cash flows may offer some the most attractive relative value. They seem priced to overly aggressive prepayment assumptions, undervaluing the embedded IO.
CRT ‘last cash flows’ are somewhat of a misnomer in that the moniker actually refers to the junior mezzanine tranche of CRT deals with sequentially tranched subordinate B classes beneath them. Given the sequential structure that the GSEs employ in their credit risk transfer programs, these mezzanine classes should tighten as the structure de-levers and the junior mezzanine or M2 class of most deals becomes the current pay bond. The structure affords investors a potentially attractive source of total return as the deals de-lever. In addition to spread compression through deleveraging, these mezzanine bonds have experienced far greater spread tightening than other mezzanine bonds like RPL BBBs which also employ a sequential structure and RMBS 2.0 BBB classes. (Exhibit 1)
Exhibit 1: CRT M2s have tightened more than other residential credit
As a result of overall spread tightening and prepayment risk, certain collateral profiles may provide better carry and total return than others. In constructing a framework for relative value, the universe of last cash flows is effectively bifurcated into two groups, last cash flows that are current pay and those that are still locked out. Fast speeds have pushed a disproportionately large amount of the last cash flow universe to either open pay or rapidly approaching opening up. In fact bonds issued as recently as the first half of 2019 are already open window. Carry associated with premium IO will likely be the largest driver of outperformance and as such OAS becomes the primary driver of relative value for the majority of the universe.
Most of the last cash flow universe trades with negative effective duration, so to establish some idea of risk and return, OAS is mapped against average life instead. Based on this framework, it appears that higher premium higher coupon M2 and M3s issued in 2016 and 2019 may offer some of the most attractive relative value. This is likely a result of mispricing these bonds to overly aggressive prepayment assumptions, which may not be realized over a longer horizon. Some examples of this are STACR HQA1 and DNA1 M3 and STACR 2019-HQA4 M2. Amherst’s proprietary long –term prepayment projections on the 2016 bonds are 12 CPR and 11 CPR respectively and 18 CPR for the 2019 STACR bond. The 2016 HQA deal is currently prepaying at a 6-month speed of 18 CPR and at a roughly $112 price, the difference between the 6-month average CPR and Amherst’s projected long term average speed equates to 150 bp difference in nominal spread. (Exhibit 2) And by contrast, the majority of 2014, 2015 and 2018 vintage bonds appear to be overvalued.
Exhibit 2: Higher coupon, higher premium last cash flows may be undervalued
Comparing OAS and prepayment projections across the las cash flow universe versus Yieldbook projections shows that proprietary prepayment projections are slower for the overwhelming majority of the universe. This is largely attributable to faster burnout in the proprietary model which appears to be in line with empirical speeds. Given this, the disparity will be more pronounced in moderately seasoned bonds that are projected to hit terminal speeds sooner and less pronounced in more highly seasoned collateral that are already burnt out.
Given almost the entirety of the universe trades at a meaningful premium to par, slower prepayment projections translate to wider OAS—in some cases substantially wider. Unsurprisingly, these differences are most pronounced in deals backed by loans that are roughly 36 WALA or greater as 2016 vintage deals show the widest OAS differential since these bonds were issued with some of the largest stated coupons against the backdrop of the meltdown in high yield in the first quarter of that year and subsequently trade at some of the highest premiums. (Exhibit 3) No model is foolproof, of course, and higher OAS can represent a premium for taking the risk that actual speeds diverge from model projections. Still, investors looking to extract relative value from open window CRT last cash flows potentially have significantly more upside to outperform on bonds with greater carry against the backdrop of slower long term speeds.
Exhibit 3: Slower speeds and bigger premiums drive wider OAS