Uncategorized
Emerging prepayment patterns in non-QM RMBS
admin | March 22, 2019
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.
Prepayments in non-QM mortgage trusts remain elevated, but not all non-QM trusts are the same. The trusts reflect an amalgamation of credit profiles, documentation and occupancy types. Borrowers with lower credit scores have prepaid relatively faster than borrowers with higher scores, loans with full documentation have prepaid faster than loans with limited documentation, and owner-occupied loans have prepaid faster than investor loans. Collateral mix looks set to be a key part of relative value in RMBS 2.0.
Faster speeds in low FICO loans
Prepayments in non-QM mortgage trusts generally have run faster than other private-label securitizations, but speeds across trusts have not been uniform. The differences reflect variations in collateral from trust to trust and shelf to shelf. Speeds tend to run somewhat slower on shelves closer to prime and faster on shelves with more non-prime (Exhibit 1). Intuition might suggest greater credit availability and less refinancing friction for prime loans, but non-prime shelves show the effect of credit curing.
Exhibit 1: Constant repayment rates across non-QM shelves by loan age
Source: Amherst Insight Labs, Amherst Pierpont Securities
Cutting prepayments by FICO and seasoning in non-QM trusts shows that lower FICO loans have by and large paid at least as fast if not faster than prime borrowers. Absolute prepayment rates accelerate on lower FICO loans as they season, a pattern attributable to credit curing as credit scores improve and past credit events recede into the past. Prepayments also may be fueled by a modest relaxation of underwriting standards in recent years. Loans with an average original FICO score of 640 to 660 with three months of seasoning have prepaid at roughly 13 CRR. With 30 months of seasoning, they prepay around 60 CRR. Loans with an original FICO score of 760 to 780 with three months of seasoning have prepaid at roughly 17 CRR. With 30 months of seasoning, they prepay around 20 CRR or below (Exhibit 2).
Exhibit 2: Constant prepayment rates by FICO score and loan age
Source: Amherst Insight Labs, Amherst Pierpont Securities
Faster speeds in full documentation loans
Non-QM trusts also contain limited document loans in varying amounts. Non-QM underwriting often uses bank statements to verify assets and income using anywhere from two years of statements to as little as one month. Additionally, non-QM underwriting can use asset depletion to qualify borrowers. Asset depletion uses a borrower’s liquid assets to qualify them for a mortgage. This is generally calculated by drawing down a borrower’s assets over a fixed period of time to determine an ability to repay the loan. The lender often makes assumptions about the borrower’s ability to sell or liquidate assets over time. Additionally, investors in rental properties can qualify based on the expected rental income that the property can generate. Although not controlling for key attributes that drive prepayments like WAC, refi incentive and loan size, fully documented loans in non-QM trusts have paid roughly 10 CRR faster than limited document loans (Exhibit 3).
Exhibit 3: Cumulative repayment rate on full and alternative documentation loans across QM
Source: Amherst Insight Labs, Amherst Pierpont Securities
After controlling for refinancing incentive—gross WAC over the PMMS rate with a six-week lag—full documentation loans have prepaid approximately 11 CRR faster than limited documentation loans. Full doc loans have generally prepaid over 32 CRR with limited document loans averaging just over 21 CRR. The largest divergences came in deep in-the-money loans where full doc loans with 3.5 points of refi incentive paid nearly 20 CRR faster than limited doc ones, and those with 4.5 points of refi incentive paid 21 CRR faster.
Compensating collateral characteristics can level the playing field when it comes to limited documentation loans in non-QM. In lower CLTV buckets, alternative document loans with enough seasoning can actually prepay faster than full doc loans, likely a result of tightening risk premiums over time on less risky, limited doc loans. Conversely, prepayment rates on higher CLTV loans are substantially higher with full documentation; full doc loans with greater than 90 CLTVs, at one extreme, have been 50 CRR faster than limited doc ones (Exhibit 4).
Exhibit 4: Constant repayment rate differences by documentation and original CLTV
Source: Amherst Insight Labs, Amherst Pierpont Securities. Note: Observations >30 WALA excluded due to limited sample across buckets.
Slower speeds in investor loans
Speeds on investor loans were significantly slower than owner occupied ones. In aggregate, investor loans have prepaid 16 CRR slower than primary residences. Controlling for refi incentive, investor loans still prepay 11 CRR slower than owner-occupied ones. Comparing investor and owner-occupied loans with similar FICOs or LTVs generally suggest that investor loans run consistently slower than owner-occupied ones even in stronger credits with less friction to refinancing. Given this, investor loans may provide attractive prepay protection in otherwise fast paying non-QM deals. (Exhibit 5)
Exhibit 5: Cumulative repayment rates by occupancy status
Source: Amherst Insight Labs, Amherst Pierpont Securities