Uncategorized

Rising negative convexity in non-QM MBS

admin | August 7, 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.

Despite elevated delinquencies, there are signs of increasing negative convexity in non-QM MBS. Prepayment speeds are starting to spike again on certain types of loans. And the low absolute level of rates, tightening liability spreads and high advance rates on new securitizations suggests rising issuer call risk in the non-QM market as well. Rising prepayments and elevated call risk should help returns on discount mezzanine and subordinate bonds backed by better performing loans. Premium bonds, however, look vulnerable to spread widening.

Speeds pick up in July

 Unlike the phenomenon occurring in prime jumbo trusts where it appears that some loans are leaving private trusts for more favorable agency execution, July’s spike in non-QM prepayments was led, at least in some part, by a surge from limited documentation loans. Controlling for a population of owner-occupied loans with balances between $400,000 and $800,000, prepayment speeds on limited documentation loans increased substantially month over month while speeds on full documentation loans fell substantially, converging at roughly 35 CRR. The phenomenon was evident in investor properties as well where full doc speeds fell below 10 CRR in July while limited document loans on investor properties paid north of 20 CRR controlling for the same loan balances as owner occupied ones. The increase in speeds on limited documentation loans likely comes as somewhat of a surprise to the market given more stringent credit overlays put in place by non-QM lenders in recent months. (Exhibit 1)

Exhibit 1: Limited doc loan prepayments surge in July

Source: Amherst Insight Labs, Amherst Pierpont Securities

Especially with regard to owner-occupied loans, the convergence in prepayment rates between full documentation and limited documentation loans may be somewhat of a temporary anomaly. A historical S-curve analysis shows that full documentation loans are significantly more sensitive to refinancing incentive than limited documentation ones after controlling for any risk-based pricing or SATO differences. Given 100 bp of SATO-adjusted refinancing incentive, full documentation owner-occupied loans tend to prepay roughly 15 CRR faster than limited documentation loans with the same amount of incentive.

That difference is far less prevalent in investor loans where given the same refinancing incentive, full documentation loans pay 4 CRR faster than limited document ones. Given the smaller difference in prepayment sensitivity across investor properties, it seems plausible that speeds on limited documentation investor loans may remain elevated relative to those on fully documented ones. The likely driver of faster in-the-money speeds on limited documentation investor loans are loans that are underwritten using asset depletion rather than the rental income that the property may generate as asset depletion loans with 100 bp of refinancing incentive have historically prepaid faster than 40 CRR while loans that were underwritten using a CMBS style debt service coverage model have only prepaid at 20 CRR given the same amount of refinancing incentive.  (Exhibit 2)

Exhibit 2: Faster speeds in limited documentation loans despite historically less sensitivity to refinancing incentive

Source: Amherst Insight Labs, Amherst Pierpont Securities

Another surprise in July’s remittance, especially in light of the anecdotal tightening of the credit box in the sector, is the spike in prepayment rates on lower FICO loans. Prepayment rates on borrowers with FICO scores between 660 and 680 rose by 25% month-over-month and are roughly in-line with rates evident prior to Covid-19-related market disruptions and subsequent fundamental credit deterioration. Rising prepayments on lower FICO loans is also somewhat surprising as delinquency and modification rates for non-QM borrowers with comparable FICOs have surged from 5% earlier this year to nearly 30%. (Exhibit 3)

Exhibit 3: Prepayments rise along with delinquencies on low FICO loans

Source: Amherst Insight Labs, Amherst Pierpont Securities

Another source of negative convexity may be returning to the non-QM market in the form of issuer calls. The drop in rates, favorable advance rates and tightening liability spreads have likely pushed the cost of issuing a non-QM deal to a cost lower than the market has seen to date and likely materially lower than those of deals issued last year when interest rates were significantly higher than they are today.  Using Starwood’s  recently priced STAR 2020-3 transaction as proxy for non-QM new issue execution shows the overall weighted average spread on liabilities issued were just over 200 bp and the issuer was able to achieve a 98% advance rate on par collateral by selling down from a ‘AAA’ through ‘B’ rated class. Using the deal’s WAL at the pricing speed of 2.89 years and the interpolated swap rate at that point on the yield curve implies that the all in cost to the issuer is in the neighborhood of 2.25% not inclusive of any other expenses associated with issuance. (Exhibit 4)

Exhibit 4: STAR 2020-3 Cost of Funds

Source: Bloomberg LP, Amherst Pierpont Securities

A significant potential deterrent to issuers calling outstanding non-QM trusts is the price of the underlying loans relative to the par strike on their call option. If loans are trading at a discount to par, there is little to no incentive for the issuer to call the deal absent a scenario where the loans are trading at a slight discount and the savings on the cost of funds more than offsets the loss on the difference between the price of the loans and par.

Despite elevated levels of delinquencies and payment forbearance in many non-QM trusts, it appears that the overwhelming majority of collateral backing non-QM trusts still is worth more than par. Pricing the loans 125 bp wider than current deal execution at a 3.5% yield to the ALIAS pay model base case fair value scenario on roughly 50 outstanding non-QM trusts shows that even against the backdrop of elevated liquidation rates, the collateral trades through par, primarily driven by low projected lost severities on loans that ultimately liquidate. (Exhibit 5)

Exhibit 5: Valuing the non-QM collateral universe

Source: Amherst Insight Labs, Amherst Pierpont Securities

Investment implications

Historically, holders of non-QM mezzanine and subordinate bonds were effectively relegated to being content with carry as the assets were generally par- priced at new issuance and would trade with little duration into a rally as a function of the call. The call also provided mezzanine and subordinate investors, even those in the sequential part of the capital structure minimal upside from roll down and deleveraging, especially in deals with shorter date calls as opposed to those with collateral balance clean up calls as the deals would trade with limited spread duration and get called if the collateral was performing well. That paradigm may have shifted to some extent as investors in deeper discount mezzanine and subordinate bonds may benefit from pulling to par as the result of the deal being called. Conversely, or more accurately similar to pre-COVID price action, investors holding premium bonds may be subject to spread widening as a result of rising prepayment speeds and negative convexity associated with the call.