By the Numbers
The changing landscape for CRT
Chris Helwig | September 24, 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 credit risk transfer market appears headed back to pre-pandemic form where both Fannie Mae and Freddie Mac distribute a steady stream of credit risk to capital markets and insurers. But recent developments suggest an increased focus on the costs of that credit protection both at issuance and over time as loans and structures de-lever. Fannie Mae’s announced plan to return to CRT, Freddie Mac’s recent tender offer and Freddie Mac’s changes to its latest deal structure all point in that direction.
Supply trending up with an eye towards keeping costs down
The current Federal Housing Finance Agency administration has already made substantial departures from policies enacted under former Director Calabria, but one place where current leadership aligns with Calabria is the costs associated with transferring credit risk. While Calabria cited the amount of premium paid by the GSEs relative to losses incurred as a condemnation of the programs, current leadership likely has a more nuanced view. Based on comments made last week, current leadership sees CRT as insurance against high-cost, low probability events. But the cost of that protection needs to be weighed against capital requirements, the cost of capital and other factors.
The first move by the enterprises to reduce costs associated with CRT was made by Freddie Mac through its tender offer on eight seasoned mezzanine classes of outstanding CRT transactions. The initial tender offer of $650 million original face was subsequently upsized to $1.629 billion. The upsize was just less than $1.4 billion in market value across the eight classes and represented nearly 70% of the total original face issued.
The tender looks beneficial to both Freddie Mac and investors. Freddie Mac was able to retire a large swath of protection that had become increasingly expensive as both loans and structures de-levered. Investors received a premium to recent trading levels for bonds with shorter spread duration and limited continued price upside. The success of the offer likely paves the way for comparable Freddie Mac offerings in the future as there is still a substantial amount of outstanding STACR issuance that carry both high implied guarantee fees and elevated loss coverage multiples. A tender from Fannie Mae seems less likely, at least in the near term. Given that Fannie Fae has continued to aggregate credit risk over roughly the past 18 months while Freddie Mac has continued to tap the CRT market, a move to further reduce outstanding protection seems improbable until Fannie Mae has transferred some of the retained risk to the capital markets or insurers.
A second move to reduce CRT costs is the introduction of a 5-year call on Freddie Mac’s recently announced STACR 2021-HQA3 transaction. The shorter, par-priced date call allows Freddie Mac to retire protection sooner than any other previously issued transaction. While the structure is still a sequential-pay four tranche deal, when run to call at the pricing speed, the M2 class is a tight eight-month principal window and the B1 and B2 classes are 5-year bullet cash flows. One of the major draws of investors to CRT is the total return associated with rolling down the spread curve as the sequential structure pays down over time. The introduction of the short-dated call trims potential upside from that roll-down, and investors will likely price in the additional negative convexity associated with the call. Freddie Mac also extended the maturity date from their prior transaction from 12.5 years to 20 years. The shortening of the call and extension of the maturity increases the negative credit convexity for investors as they will likely get called out at par if the collateral performs well and extend if there are outsized delinquencies, pool losses or slower deleveraging of the structure from muted prepayments on lower WAC loans.
CRT issuance may look different in other ways going forward.
Fannie Mae’s announced plan to get back into CRT stated the agency expect deals to be structurally similar to prior offerings. But it seems possible that if the new STACR structure meets with substantial investor demand, then Fannie Mae could follow suit. In terms of the projected size of fourth quarter issuance, Fannie Mae announced it may issue as many as four deals in the fourth quarter starting in mid-October but may to not issue in all four periods put forth in their announcement. Historically, Fannie Mae’s deals have ranged from $1 billion to $1.5 billion and there is little evidence to suggest the agency would divert materially from that, potentially putting their fourth quarter issuance anywhere from $4 billion to $6 billion.
One potential interesting wrinkle could come if Fannie Mae issued deals backed by its seasoned retained guarantee book, especially is the agency issued a thinner slice of risk similar to past ‘Seasoned B Tranche’ or SBT transactions. Assuming $1.2 trillion of exposure with no existing credit enhancement, roughly 25% of that or $300 billion could be close to 24-months seasoned and highly de-levered due to nearly unprecedented, broad based home price appreciation over the past two years. If the agency were to issue a 50 bp slice of risk off this deal, it could target a single issuance of $1.5 billion referencing late 2019 and early 2020 production. Given the thinness of the tranche, the agency could set the minimum credit enhancement test well above the initial detachment point, at 1.0%-1.5%, effectively creating a synthetic mezzanine class above the issued bonds. The benefits of this would allow the structure to de-lever giving Fannie Mae more credit protection and would provide investors with some lock-out and spread duration on loans that are well in-the-money and likely to prepay fast.
Additionally, it appears incremental potential economic risk may be passed on to CRT bondholders in the future, although it would likely take another episode of large-scale extension of payment forbearance to trigger meaningful trust losses. The new Freddie Mac deal introduces an allocation of losses associated with payment deferrals. In their previous 2021-DNA5 transaction, Freddie Mac removed the coupon margin on the retained B3 class of the deal. Prior to that transaction, the B3 carried a substantial coupon that was first in-line in the loss allocation waterfall to absorb losses resultant from interest rate modifications. In this transaction, rate modifications in excess of the SOFR+0 coupon will be allocated to the B3 principal balance, consistent with previous deals. Additionally, losses associated with payment deferrals will be passed on to the B3 class. The largest source of potential losses would be interest advances made by Freddie Mac on loans during a forbearance period that receive a payment deferral at the end of forbearance. It appears these losses will be allocated to the deal cumulatively once the deferral is granted at the loan level.
Under forbearance policies put in place during pandemic, servicers have to make advances for four months. The obligation to pay interest on loans in forbearance in MBS pools or CMOs beyond four months falls back to the GSEs. In the case of a borrower exiting forbearance after eighteen months, the GSEs would have paid out fourteen months of interest, which, under a deferral plan, would be capitalized to the maturity date of the loan. However, the probability of this change translating to meaningful losses appears small. For example, if 6.0% of a pool with an average WAC of 3.5% went into forbearance and 50% of those borrowers ultimately received a deferral that would translate to just over 400 bp of interest advances over and above the servicer obligation or roughly 12 bp of net losses. This potential loss would have to be probability-weighted and even assigning a one-in-ten chance that there is another large-scale forbearance plan enacted during the life of the transaction that only amounts to just over 1 bp of expected pool losses.