By the Numbers
Prepayment protection in 0% VA pools
Brian Landy, CFA | January 15, 2021
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.
Loans guaranteed by the Veterans Administration are known to prepay much faster than other government-insured loans, and custom Ginnie Mae pools that exclude VA loans can offer significant protection against prepayment risk. Ginnie Mae multi-issuer pools with lower coupons are dominated by VA loans since VA borrowers typically receive lower interest rates than other borrowers. Most G2SF 2.0% pools contain around 80% VA. But the market is currently offering 0% VA pools with very little pay-up. They appear to be one of the market’s most inexpensive sources of prepayment protection.
A review of prepayment speeds during the most recent refinancing wave illustrates that VA loans continued to prepay much faster than FHA loans (Exhibit 1). The pattern is clear in S-curves for FHA and VA loans delivered into multi-issuer pools with between six months and 36 months of seasoning from March through December 2020. The average S-curve for all loans in the multi-pool is included as well. The S-curve includes only voluntary prepayments and excludes buyouts to eliminate distortions from the large bank buyouts last summer. Only FHA loans greater than $225,000 original loan size are included in the FHA-only S-curve, since smaller loans ought to be sold into low loan balance pools. This paints a more conservative view of what a 0% VA pool would look like. However, many of these loans have been sold into the multi pools.
Exhibit 1: VA loans prepaid much faster than FHA loans during the pandemic
The multi pool S-curve, however, is not indicative of how a specific multi pool will prepay. Prepayment rates on pools with lower refinancing incentives come from lower coupon pools with heavier VA concentrations, while the higher incentives come from higher coupon pools with lower VA concentrations. This causes the multi-pool S-curve to converge towards the FHA S-curve at higher incentives.
An investor buying a multi pool will get one with a specific VA concentration. Most 2.0% pools are roughly 80% VA, while most outstanding 2.5% pools are 50% to 70% VA. Investors should assume the G2SF 2.0% TBA is 80% VA and the 2.5% TBA is 70% VA. Similarly, most outstanding 3.0% multi pools are 40% to 50% VA, so investors should assume 50% VA for the TBA.
Prepayment differences between a likely multi pool stand out in a picture showing the same FHA and VA S-curves and a blended S-curve assuming 80% VA, which is a proxy for the 2.0% TBA pool (Exhibit 2). The blended S-curve eliminates most of the skew across rate incentives seen in the previous chart due to the changing VA share. The FHA loans should refinance 40% slower than the multi pool; illustrated by the red dotted line where the refi response is dialed down to 40% of the blue baseline. The high VA concentration suggests that the 2.0% TBA would have extremely fast speeds if interest rates were to drop.
Exhibit 2: A 2.0% pool with no VA loans should refinance 40% slower than TBA.
The dialed S-curve meets the FHA loans at the highest incentive but is still somewhat faster than FHA at lower incentives. This is because the VA loans tend to be a higher loan size at those incentives. The multiplier was chosen to give a more conservative view of the speed benefit of the FHA loans.
Similarly, the 2.5% TBA is modeled using a 70% VA share (Exhibit 3). This TBA should also exhibit significant negative convexity, and a 0% VA pool should refinance 35% slower than the TBA.
Exhibit 3: A 2.5% pool with no VA loans should refinance 35% slower than TBA.
Similarly, the 3.0% TBA is modeled using a 50% VA share (Exhibit 4). This places the TBA S-curve halfway between the FHA and VA S-curves, and a 0% VA pool should refinance 30% slower than the TBA.
Exhibit 4: A 3.0% pool with no VA loans should refinance 30% slower than TBA.
The theoretical pay-ups implied by these speed differences is substantial. Pay-ups are calculated by running a recent production multi pool from each coupon through Yield Book at the TBA price to get an OAS. The multi pool is used as a proxy for TBA; these OASs are much lower than Yield Book’s TBA OASs, which suggests these are more representative than Yield Book’s assumption of what might be delivered into a TBA contract. Each bond is run a second time at the TBA OAS, through a version of the model dialed to have slower refinance speeds expected of a 0% VA pool. The price difference is the theoretical pay-up.
Exhibit 5: Theoretical pay-ups range from $1-18+ to $2-20+ yet very few pools are being made
The largest pay-up is $2-20+ for a 0% VA 2.5% pool. But there has been almost no production or trading of these pools in this coupon. There were some recent trades in the 2.0% coupon, with pay-ups averaging 3/32s. That is far below the theoretical pay-up of $1-18+. The 2.5% coupon benefits the most since it is projected to have extremely fast prepayment speeds and also has an extremely high VA share.
Investors might be concerned about the buyout risk posed by FHA loans, which could lower the pay-up. But this risk should be much lower in new production pools since the loans were originated well after the onset of the pandemic. Of course, the economy could weaken and raise buyout risk. Another consideration is the Ginnie Mae rolls in these coupons are trading special, as are the UMBS rolls in 2.0%s and 2.5%s. But these pay-ups are much higher than the carry advantage in the roll, so the roll would need to remain special for a long time to erode the pay-up. Finally, 0% VA pools come out through the Ginnie Mae custom program and are not eligible for TBA delivery, lowering their liquidity and likely showing up in lower marks from pricing services. Nevertheless, portfolios able to bear the liquidity difference a pricing risk should still see value in the fundamentally more stable likely cash flows of 0% VA pools.