Scanning for convexity in prime jumbo
admin | October 9, 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.
Prepayment speeds in prime jumbo show no sign of slowing down. And traditional sources of convexity such as seasoning and loan size have offered little to no protection with prepayments on more seasoned trusts and loans with lower balances actually coming in faster than others. Safe harbor may be in more nuanced stories such as higher LTVs loans or increased concentration to particular servicers.
Finding prepayment protection in the prime jumbo market has become increasingly challenging as sustained low rates have continued to drive speeds faster. The issue is compounded by elevated speeds in loans that investors may assume should afford them some refuge from faster prepayments. Jumbo conforming loans are prepaying faster than those with non-conforming balances. And seasoned trusts are exhibiting faster speeds than more recent vintages, partially driven by greater concentrations of lower balance loans (Exhibit 1).
Exhibit 1: Seasoning offers little prepayment protection in prime jumbo
Conforming jumbo borrowers in private label trusts with 100 bp of refinancing incentive have prepaid 10 CRR faster than their non-conforming jumbo counterparts at 53 CRR and 43 CRR, respectively, and exhibit steeper S-curves than both traditional conforming balance loans and non-conforming ones. Elevated speeds on jumbo conforming collateral are likely driven by both the decoupling between conforming and non-conforming rates and the generally lower friction to refinancing associated with agency-eligible loans. It appears both these drivers may persist for the foreseeable future and speeds on jumbo conforming loans will likely remain elevated as a result.
Elevated jumbo conforming speeds appear, in some part, to be driving faster speeds in seasoned collateral. As jumbo loans in seasoned trusts have paid down through amortization and curtailment, the population of jumbo conforming loans in these trusts is much larger than those of newer vintage deals. As of September, the 2017 vintage has the largest population of loans with balances between $400,000 and $600,000, making up 42% of all loans in the cohort. By comparison, loans with those balances make up just 25% of the 2019 and 2020 vintages. Conforming jumbo loans from 2017 prepaid at 58 CRR last month while similar loans in the 2018 vintage paid 61 CRR. By comparison, loans with balances greater than $1 million prepaid at 35 CRR and 30 CRR in the respective cohorts (Exhibit 2).
Exhibit 2: Jumbo conforming speeds remain elevated across vintages
Finding protection in LTV and servicer stories
Given the lack of prepayment protection afforded by loan balance and seasoning, investors may have to look to more nuanced stories to find some protection from continued elevated prepayments. One potential source of protection is higher LTV loans which exhibit significantly flatter S-curves than those of lower LTV ones. Given the higher risk-based pricing associated with higher LTV loans, that incremental spread is captured and controlled for when evaluating sensitivity to different amounts of refinancing incentive. After controlling for this spread or, SATO adjustment, it appears that given the same amount of refinancing incentive, high LTV loans prepay much slower than lower LTV loans. At 50 bp of refinancing incentive, loans with original combined LTVs of 90 or greater prepay at roughly 5 CRR while loans with lower LTVs prepay at 35 CRR or faster. (Exhibit 3) As the incentive goes further in the money the relationship is roughly similar as the high LTV cohort prepays at roughly 17 CRR given 100 bp of incentive while 70-80 LTV loans prepay at roughly 50 CRR.
Exhibit 3: High LTV loans exhibit flat S-curves
It seems plausible that dampened prepayments may be due to higher LTV borrowers having fewer channels to refinance their loans relative to lower LTV ones and may be a material friction to refinancing. Additionally, while SATO has been controlled for, there may be other attributes to higher LTV loans that may drive slower speeds as the analysis does not control for other attributes. However, after controlling for loan size, it appears that high LTV loans still offer significant prepay protection versus lower ones in both jumbo conforming and non-conforming cohorts.
Certain prime servicers have also exhibited flatter S-curves than others. Conduit transactions will have multiple seller/servicers, and as a result, gaining isolated exposure to slower servicers may be somewhat challenging. But investors can add convexity by adding exposure through pools with larger proportions of slower servicers. One major prime servicer that stands out is First Republic. After controlling for SATO and looking solely at fixed-rate originations, First Republic originated and serviced loans prepay at roughly half the speed of some other major prime seller servicers given 50 bp of refinancing incentive. At 50 bp of incentive loans serviced by both First Republic and Wells Fargo have prepaid at an average rate of 22 CRR. However, as the option goes deeper in-the-money, speeds on First Republic serviced loans remain relatively muted at 35 CRR while, Wells Fargo serviced loans have prepaid upwards of 57 CRR given 100 bp of refinancing incentive (Exhibit 4).
Exhibit 4: Prepayment S-curves by seller/servicer
Breaking the analysis into jumbo conforming and non-conforming buckets, jumbo conforming loans serviced by First Republic have prepaid at just over 17 CRR given 100 bp of incentive while loans with comparable incentive serviced by United Shore have prepaid upwards of 53 CRR and those serviced by Loan Depot have prepaid at 58 CRR. Looking at non-conforming balances, First Republic still maintains slower in-the-money speeds, albeit more modestly, than all other servicers as loans with 100 bp of incentive have prepaid at roughly 35 CRR. Those serviced by JP Morgan Chase with similar incentive have prepaid at 45 CRR, while those serviced by Flagstar have prepaid upwards of 50 CRR.
Thoughts on relative value
Investors can get relatively concentrated exposure to the convexity afforded by high LTV collateral as certain shelves like JPMMT issue deals backed solely by high LTV loans. However these deals are not backed solely by loans with greater than 90 LTV ratios and, as a result, in-the-money speeds should be somewhat suppressed by the population of those loans but still somewhat elevated given the material concentration of loans with LTVs greater than 75 but less than 90 in these pools. Despite the improved convexity of higher LTV loans, pass-throughs backed by this collateral actually trade behind generic jumbo pass throughs. 3.0% LTV pass throughs currently trade at $1-28 back of UMBS 3.0%, a roughly 4/32 concession to generic 3.0% pass throughs. Admittedly, some of the convexity associated with the collateral is mitigated to some extent by the structure as LTV pass throughs often have double the amount of credit enhancement as their generic counterparts making the LTV pass through more sequential-like given the shifting interest structure. Even with the greater structural leverage, it appears that the convexity associated with higher LTV collateral should trade at premium not a concession to more negatively convex generic bonds.