By the Numbers
Assessing prepayment and default risk of PLS servicers
Brian Landy, CFA | July 15, 2022
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 agency MBS market has long recognized that a servicer’s business practices influence prepayment speeds, and pools from different servicers trade at different prices. Amherst Pierpont’s monthly agency servicer prepayment rankings help identify fast and slow servicers. The same approach applies to the effect servicers have on the prepayment speeds, delinquency rates and value of private-label MBS. Analysis of PLS shows that not all servicers were created, or, at least operate, equally.
The PLS prepayment and default ranking reports can be found at the APS Portfolio Strategy website in the Library under APS Reports. There are reports that rank PLS servicers by prepayments and delinquencies. There are also reports that rank PLS shelves by prepayments and delinquencies.
Measuring servicer prepayment risk
Directly comparing the prepayment speeds of servicers can be misleading. A servicer’s business might be concentrated in a state with a higher jumbo conforming limit, so its non-agency business has a higher average loan size than a servicer that focuses on states without any high-cost areas. It would be unfair to directly compare those two servicers’ speeds without accounting for the expected prepayment difference from loan size. Similar issues can arise around note rate, loan age, and other collateral attributes. Product type also matters—for example, a non-QM loan might prepay differently than a jumbo loan.
A good measurement of a servicer’s effect on prepayment speeds needs to control for collateral characteristics. Amherst Pierpont’s PLS servicer rankings calculate two prepayment speeds for each servicer—its actual prepayment speed (“CPR”) and a reference speed (“Ref”) that controls for differences in collateral characteristics (Exhibit 1). The reference speed is calculated using the actual prepayment speeds across the entire universe of prime jumbo prepayment speeds but is adjusted to match that servicer’s portfolio composition. That explains why the reference speed is faster for some servicers and slower for others—it encapsulates the expected prepayment speed of that servicer’s loans.
Exhibit 1. Prime jumbo servicer prepayment ranking
As of June 2022 remittances, for a 3-month horizon.
Source: CoreLogic, Amherst Pierpont Securities
The column headed “%ΔRef” shows how much faster, or slower, a servicer’s loans prepaid than expected. This is the effect a servicer has on prepayments after controlling for collateral characteristics. The servicers are sorted by this column, which is highlighted. For example, PennyMac’s loans prepaid at 8.6 CPR, which is 4.5% faster than the expected 8.3 CPR. The column is calculated using SMM, not CPR, because using SMM is more accurate.
This snapshot shows that loanDepot’s loans prepaid 9.9% faster than expected and is one of the faster servicers in the report. On the other hand, Chase is closer to neutral, only 2.2% faster than reference. United Shore was also close to neutral despite a reputation for fast prepayment speeds in agency MBS. However, its agency speeds have moderated over the last year.
Ranking servicers by delinquency rates
A comparable report is also available that ranks servicers by the percentage of their portfolio that is at least 60 days delinquent (Exhibit 2). The table shows delinquency rates at the end of 2020 when many loans were not paying while in COVID-19 forbearance. For example, 4.9% of Quicken’s servicing in prime jumbo deals were at least 60-days delinquent at the end of 2020. That was 23.2% greater than the expected delinquency rate of 3.9%.
Exhibit 2. Prime jumbo servicer delinquency rate ranking
As of January 2021 remittances, for a 3-month horizon.
Source: CoreLogic, Amherst Pierpont Securities
Shelf reports and other notes
Reports are also available that rank different shelves by prepayments and defaults. The loan types covered include:
- PLS: prime jumbo, investor, and expanded prime
- Non-QM: bank statement, DSCR/asset depletion
- Seasoned RPL
Methodology
The reference metric—speed or delinquency rate—is calculated by grouping loans based on loan type, note rate, loan age, loan size, credit score, LTV, SATO, state, and performance month. Each loan in a servicer’s portfolio is assigned a speed from the appropriate reference bucket, and the weighted average actual and reference speed is calculated for that servicer.
A separate write-up for the agency report includes a demonstration of the methodology and how it is calculated. The example is simplified but accurately shows the process used to generate the rankings. It is comparable to using a non-parametric model with interactions.
The full report ranks servicers over the last 3-, 6-, and 12-month horizons. With larger servicers this can help show if there has been a trend in behavior. For smaller servicers using a longer horizon minimizes noise from small sample size. Within each horizon the servicer closest to neutral is highlighted. A version of the report is also available that ranks deal shelves instead of servicer using the same methodology.