Uncategorized
Comparing jumbo loans across agency and private MBS
admin | August 24, 2018
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.
Originators have securitized high balance conforming loans through private-label MBS in increasing amounts this year. Investors consequently can get exposure to these loans through a range of securities. A side-by-side comparison of agency and non-agency high balance conforming loans with similar attributes highlights noticeable differences in prepayments.
The particulars:
- Agency pools tend to pay faster than comparable WAC and WALA private-label loans – generally attributed to higher loan balances
- PLS loans appear to ramp faster than comparable agency pools
- ‘Near miss’-prime conforming jumbo loans tend to ramp fast but burn out quickly
- In certain seasoned transactions, conforming jumbo loans are paying faster than larger balance non-conforming loans
An increasing flow into private-label MBS
The population of high-balance agency-eligible loans in deals is growing and getting increased focus from both originators and investors. However, the securitization of conforming jumbo loans in private-label deals is by no means new. It may be widely held that the first post-crisis incidence of agency-eligible loans in PLS deals occurred in the well-conceived but ill-fated CHASE private risk transfer securitizations in 2016. But conforming jumbo loans made up a meaningful share of loans in private-label trusts a year before those deals. JPMMT 2015-3 and 2015-4 were comprised of 25% and 19% conforming jumbo collateral, respectively, at issuance. Deals going back as far as 2014 have been comprised of more than 10% conforming jumbo loans – further illustrating that private-label securitization of agency-eligible loans is by no means novel.
Exhibit 1: The population of agency eligible loans in private-label trusts
Source: Amherst Insight Labs, Amherst Pierpont
Given rather pristine credit across agency and private-label jumbo collateral, this analysis focuses on prepayment behavior of like borrowers in Fannie Mae jumbo pools and private-label trusts. The analysis focuses on three major prime jumbo shelves, CHASE, JPMMT and SEMT, uses deals with only 100% fixed-rate collateral and defines conforming jumbo loans as having an original balance between $417,000 and $625,000 and is comprised of two years of prepay history. The selection criteria for these loans are a simplifying assumption. As home prices have risen over time, conforming loan limits have risen as well and currently sit at $629,650. However, the upper bound of conforming jumbo may have little influence on the analysis. One of the first notable observations about how private-label loans stack up against their agency counterparts is that average loan balances securitized into PLS deals are by and large materially lower than their agency counterparts.
The analysis compares the conforming jumbo balances across the three shelves to their closest net WAC and WALA Fannie Mae jumbo ‘CK’ counterpart. It excludes deals where the difference in average net WAC is greater than 10 bp or difference in average WALA varies more than six months from the agency cohort. Based on this methodology, the average loan size of CK pools is roughly $71,000 greater than those of comparable loans securitized into PLS trusts. Some of the largest differences are in higher WAC CK pools relative to loans in Seqoia’s CH ‘near- miss’ prime jumbo program, where average loan balances can diverge by as much as $100,000 compared to their agency counterparts.
Exhibit 2: PLS jumbos tend to have lower average loan balances than agency
Source: Amherst Insight Labs, Amherst Pierpont
Prepayment results
Comparing loans in PLS trusts with at least two years of prepayment history show that PLS loans generally tend to prepay slower than their agency counterparts, likely attributable in large part to differences in average loan size. While generally slower, non-agency prepayments appear to be more volatile with large month-to-month swings in speeds. These lumpy prepayments can cause non-agency loans to prepay significantly faster than their agency counterparts in certain periods.
Exhibit 3: Comparing speeds across select cohorts
Source: Amherst Insight Labs, Amherst Pierpont
While the analysis attempts to control for both seasoning and net WAC to minimize the impact of both ramping and refinancing incentive, those controls are applied on an average basis. As a result, differences in prepayment speeds can be driven by differences in how far loans are up their relative ramps or by WAC dispersion within a given cohort. It nevertheless appears that low WALA non-agency loans prepay faster out of the box than their agency counterparts. Conforming balance loans in the first three JPMMT deals of this year have prepaid faster than comparable WAC and WALA Fannie Mae CK 3.5%s. One WALA conforming balance loans in those trusts have prepaid at 5-7 CPR out of the box while comparable agency pools paid less than 2 CPR at the same point on the ramp.
Figure 4: New PLS loans have prepaid faster than comparable agency pools
Source: eMBS, Amherst Insight Labs, Amherst Pierpont
Conforming balance loans in ‘near miss’ prime transactions by comparison have ramped even faster. Looking at prepays on conforming balance loans in Sequoia’s ‘CH’ issuance shows speeds north of 25 CPR post issuance with the 2017-CH1 deal peaking at nearly 40 CPR in February. These deals have paid fast but appear to burn out fairly quickly, with speeds falling from 30 CPR to the mid-single digits after a handful of months. Prepayment speeds have been elevated in all flavors of non-QM issuance, generally spurred by an overall tightening of spreads over conforming rates, creating significant refinancing incentive for borrowers despite a rise in the absolute level of conforming rates relative to when those loans were originated. Given this, there would need to be a continued tightening of risk premiums for prepayment speeds to remain elevated on out-of-the-money non-QM loans. Given the fact that ‘near miss’ jumbo conforming loans trade to relatively tight spreads over prevailing conforming rates already, an increased widening of the credit box or tightening of spreads may have a negligible impact on speeds.
Figure 5: Near-miss conforming jumbo loans pay fast but burn out quick
Source: Amherst Insight Labs, Amherst Pierpont
Interestingly, aggregate prepayments on conforming jumbo loans in certain seasoned transactions are faster than those of higher balance non-conforming loans in the deal. These loans not only had lower balances but in most cases actually had lower WACs than their non-conforming counterparts in the same deal. For example, lifetime speeds on conforming jumbo loans in SEMT 2014-3 are 23 CPR. Higher balance loans in the same deal have paid 18 CPR. The conforming jumbo loans have an average balance of roughly 500,000, while the non-conforming average loan balances are 840,000. The WAC differential on the two cohorts is only 6 bp with the conforming jumbo loans carrying an average WAC of 4.24% and the non-conforming loans a 4.18% putting both pools out of the money compared to today’s conforming rates, potentially pointing to higher rates of turnover on conforming jumbo loans relative to comparable WAC and WALA non-conforming loans. Given this, investors may look to add discount seasoned bonds backed by pools with higher percentages of conforming jumbo loans.
Figure 6: Conforming jumbo loans pay faster than non-conforming in some deals
Source: Amherst Insight Labs, Amherst Pierpont