By the Numbers

Upside in rising Ginnie Mae project loan prepayment speeds

Mary Beth Fisher, PhD | January 19, 2024

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. This material does not constitute research.

Ginnie Mae project loans continue to linger wide of other ‘AAA’ CMBS despite having both a financing advantage and zero risk-weight for banks and insurers. Reasons include a current pull-back by bank and insurance companies trying to right-size CMBS exposure, and fears that prepayment rates could remain stuck at 2% a year for recent vintages, significantly extending duration and lowering returns. But analysis of historical average prepayment speeds over more than 20 years shows that borrower prepayments rise steadily as penalties decline, independent of refinance incentive. Assessing relative value using historical speed vectors makes a compelling case for deep discount project loan securities.

It helps to compare some recent vintage Ginnie Mae project loan (GNPL) securities under the market standard 15 CPJ scenario, and a synthetic vector created from historical average voluntary prepayment speeds (CRR) since 2002 plus 100% of the standard project loan default curve (PLD) for involuntary prepayments (Exhibit 1). Although the 15 CPJ scenario is the market standard for pricing, investors run more nuanced scenarios when evaluating projected returns. But expecting prepay rates to remain stuck at current levels is as unrealistic—and contrary to historical average speeds—as presuming a static 15% prepayment curve.

Based on historical average speeds derived from declining prepayment penalties only, regardless of refinancing incentive, projected yields for deeply discounted bonds can be higher than those at 15 CPJ, even though the historical vector takes six years to rise from 0 to 20 CRR. The 7- to 8-year weighted average life (WAL) M series bonds in the table have projected returns of 6.90% to 7.50% at historical speeds, compared to 6.60% to 6.97% at 15 CPJ. Even the shorter 3- to 5-year WAL front sequentials have projected returns of 5.28% to over 6.00% at historical speeds.

Exhibit 1: GNPL bond comparative yields across prepay scenarios

Note: Indicative levels only. Project loan commitment (PLC) rate for $102 price loan. Data as of 1/18/2024. 15 CPJ is equivalent to 15 CRR (voluntary prepayments)  plus 100 PLD (the project loan default curve). The historical average CRR vector is shown in Exhibit 3.
Source: Bloomberg, Santander US Capital Markets

Investors have good reason to be wary: currently rock-bottom voluntary prepayment rates, if they persisted for a decade, lowers projected returns of even deeply discounted bonds. Prepayment speeds across all outstanding GNPL collateral hovered between 1 CPR to 3 CPR throughout 2023, meaning roughly 2% of loans prepaid last year. But these speeds are slow in part because the GNPL universe, somewhat like the single-family universe, is heavily concentrated in 2020 through 2023 production. These recent vintage loans still have the heaviest 8% to 10% prepayment penalties attached.

Moreover, delinquency rates in GNPL have been rising. Defaults in 2023 were still quite rare, but 1.1% of GNPL collateral is currently delinquent, and that proportion is expected to double by the end of 2024. Given that investors face no credit risk from fully government guaranteed loans, these defaults eventually pull through as prepayments, raising returns in deep discounted securities.

The historical average speeds represent over 20 years of data, when loans at any point faced a variety of refinance incentives and interest rate environments. Prepay speeds still exhibit a strong tendency to rise as prepayment penalties decline, then burn out as the open period extends. This suggests that some 2020 through 2022 GNPL borrowers, whose loans may be deeply out of the money to refinance, are likely to begin prepaying those loans as their penalties decline. Speeds for those vintage loans may not be 20 CRR in 2026 to 2028, but it is extremely unlikely that they will remain below 5 CRR. Property owners will likely want to tap equity, sell properties or reinvest in newer buildings. This should normalize longer-term prepay speeds, though those vintages with very low coupon loans will probably stay below historical averages.

Cliff notes on prepayment scenarios using long term historical averages

Voluntary prepayment speeds are primarily driven by the interplay of two different factors:

• The relative attractiveness of the refinance incentive; and
• The cost of the prepayment penalty, if any.

Of course, prepayment models also incorporate a lot of additional variables to project prepayment speeds. These variables can include property price appreciation, seasoning, whether or not the borrower has passed up a prior opportunity to refinance and whether the loan is a converted construction loan, among others. But looking at long term historical average speeds across refi incentives and prepay penalties separately, allows investors to somewhat isolate the very different prepay speed scenarios these factors give rise to. This can provide a baseline for projecting speeds going forward.

Impact of refinance incentive

The largest single driver of prepayment speeds, by far, is the refinance incentive. Over long periods of time, looking at prepay speeds based on refinance incentives alone, average speeds don’t materially rise above 5 CRR until the refinance incentive is 50 bp or higher, with maximum average speeds of 35 CRR to 40 CRR for refi incentives above 200 bp (Exhibit 2).

Over shorter periods of time, in unusual or extreme interest rate environments, speeds are a bit less predictable:

• In 2021, when mortgage rates hit historic lows, GNPL prepay speeds averaged 40 to 55 CRR for refinance incentives from 100 to 400 bp. These speeds may arguably have been lifted by the accelerated property price appreciation and high commercial real estate transaction volume that occurred during the time period.
• In 2023, as the Fed completed 525 bp of hikes and long term interest rates rose markedly, weighted average GNPL prepayment speeds stayed mostly between 1 to 3 CRR. Still, loans with negative refinance incentives did prepay at speeds from 3 to 4 CRR, which was slightly above long term historical averages of 0 to 2 CRR. There were also some very seasoned loans with positive refinance incentives that finally prepaid. The lower than normal speeds for these loans is possibly a reflection of burnout, given that so much of the universe of outstanding GNPLs had refinanced during the prior 2 years.

Exhibit 2: Prepay speeds accelerate quickly with rate incentives

Note: Historical S-curves are weighted averages of GNPL prepayment rates for various refinance incentives. Long term historical average is 2002 through 2022.
Source: Ginnie Mae, Santander US Capital Markets 

Impact of prepayment penalties

The S-curves don’t provide investors with a lot of information about how prepay speeds are likely to behave if interest rates are relatively static. Consider, for instance, if project loan commitment rates were to stay within a 100 bp range of current rates for the next fiveyears or longer. Then the majority of the universe of outstanding GNPLs would probably have their prepayment behavior as much, or more heavily influenced, by the decline in their prepayment penalties than any modest refinance incentive.

The most common prepayment penalty scheme for GNPLs is one where the penalty is 10% of the outstanding balance in the first year after origination, then declines by 1% per year until year 10, when the penalty is 1%. After 10 years, the loan enters an open period where it can prepay at any time without penalty. This is typically represented as a penalty string, in the format:

• 10%(12), 9%(12), 8%(12), … 1%(12), O(300)

where the number in parenthesis is the number of months the penalty applies at issuance or has remaining. Construction loans are locked out from prepayment during the construction period, which is signified by, e.g. L(24) at the beginning of the string if the construction period is expected to be 2 years.

Exhibit 3: Historical average prepayment speeds rise as penalties decline

Note: Historical speeds are weighted averages of GNPL prepayment rates since 2002. Data through January 2024.
Source: Ginnie Mae, Santander US Capital Markets

The impact of the decline in prepayment penalties is shown in Exhibit 3, which shows historical average prepayment speeds plotted against years since issuance (blue dots). The red line is a synthetic prepay vector that follows these historical trends:

• Prepay speeds tend to start out very low when penalties are still high, and climb linearly as penalties decline until about year 6, when speeds mostly flatten out at 20 CRR.
• By year 6, the prepay penalty has typically declined to 4% of the outstanding balance.
• There is a jump to 25 CRR in year 10 post issuance, when there is typically either 0 or 1 penalty point remaining.
• After year 11, when all loans are in the open period, prepay speeds begin to decline again. The few loans that tend to still be outstanding prepay a bit erratically but slowly and speeds fall back towards the low single digits.

The synthetic vector can be used to run prepay scenarios for GNPL securities when interest rates are expected to be roughly static over the near term.