Little benefit from removing VA high LTV cash-out refis
admin | May 10, 2019
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.
Ginnie Mae continues to focus on fast prepayment speeds in VA loans, which is encouraging for investors. Recently Ginnie Mae asked for input on possibly limiting or excluding certain VA cash-out refinances from multi-lender pools since these VA loans have prepaid extremely quickly. While this change helps investors, recent prepayment history suggests any improvement will be very small.
Ginnie Mae is suggesting that VA cash-out refinances with original LTV greater than 90% should be excluded from the multi-lender pool program or possibly subject to pooling limits. This is based on an analysis of prepayment speeds from January 2017 through March 2019.
The FHA program along with Fannie Mae and Freddie Mac has LTV caps on cash-out refinances, so it is reasonable to consider why the VA should be different. Congress raised the VA cash-out limit in 2008 from 90 LTV to 100 LTV. The VA appears to believe their program should follow congressional intent, while Ginnie Mae has to consider whether including unusual loans in multi-lender pools is reasonable.
The multi pools won’t slow much
If Ginnie Mae had excluded high LTV VA cash-out refinances from pools in the past, the impact on speeds would have been small (Exhibit 1). It is worth considering the impact in three different time frames:
- Performance in 2016, which is before Ginnie Mae’s first attempt to curb excessive VA prepayment speeds.
- Performance in 2017 and part of 2018, which is after Ginnie Mae began to restrict VA refinances but before additional congressional legislation enacted in May 2018.
- Performance after the May 2018 legislation.
Exhibit 1: Removing high LTV cash-out refis hardly improves speeds
Almost every cohort would have slowed less than 1.0 CPR if Ginnie Mae had excluded the VA loans, and most would have slowed far less than that level. Part of the reason is that these loans do not make up a large share of overall production, so even if they prepay significantly worse, removing them has little effect. Only about 4% of the 2018 multi-lender pools, for example, were VA cash-out refinances with LTV greater than 90%.
Potential improvement in speeds has declined following Ginnie Mae and congressional measures to improve VA prepayment speeds; all but one cohort would slow less than 0.3 CPR.
Speeds have already improved
Since enacting the May 2018 legislation, the high LTV VA cash-out loans don’t appear to prepay much worse than other VA cash-out loans. Prepayment seasoning ramps comparing prepayment speeds of VA cash-out refinances to rate/term refinances, streamlined refinances, and purchase loans only shows clear speed differences for purchases (Exhibit 2). All loans are less than 50 bp in-the-money with performance beginning in June 2018.
Exhibit 2: All VA refis have similar seasoning ramps
Purchase loans clearly prepay much slower than the various refinance loans, but cash-out refinances are not terribly different than the other refinances. In fact initially they seem to prepay a little slower, but do spike a little faster when speeds ramp up after seven months.
Separating the cash-out refinances by original LTV suggests very little difference in prepayment behavior based on the LTV (Exhibit 3).
Exhibit 3: Original LTV hardly affects cash-out seasoning ramps
None of the seasoning ramps suggest that high LTV cash-out refinances, or cash-out refinances in general, have been prepaying unusually relative to other VA loans.
Cash-out S-Curves are consistent with other VA refinance S-Curves
VA S-curves support the observation that cash-out refinances prepay at a pace similar to other refis. It is VA streamlined refinances, not the cash-outs, that stand out with a significantly steeper S-curve (Exhibit 4).
Exhibit 4: Only VA streamlines refis have steep S-curves
Comparing S-curves for VA cash-out refinances by original LTV, the S-curves for loans with original LTV greater than 90% do appear to be a little steeper, but the difference is fairly small (Exhibit 5).
Exhibit 5: Higher original LTV marginally steepens the S-curve
A different reason Ginnie Mae might target cash-out refinances
The May 2018 legislation declined to place constraints on VA cash-out refinances, instead requiring the VA to conduct a study and provide rulemaking to handle these loans. These rules became effective in February 2019, but in the meantime lenders theoretically could have bypassed the net tangible benefit and fee recoupment tests by doing a cash-out refinance. However, this loan would cost more to originate than a streamlined refinance, which should have reduced the appeal.
Ginnie Mae might be concerned that the cash-out program was elevating speeds across all VA loans, not that these loans necessarily prepay faster themselves. Making these loans more expensive would therefore slow all VA speeds. Unfortunately the disclosure data is insufficient to study this idea. And it is possible that the VA’s recent rules will be sufficient to eliminate any abuse of the cash-out refinance program.
How should these loans be handled?
Ginnie Mae has proposed that these loans could be excluded from the multi-lender pool program entirely. They would be permitted in custom pools, and possibly in a new, TBA-ineligible, multi-lender pool type to hold these and other loans expected to prepay rapidly shortly after origination. An alternative is to continue permit these loans to be included in the multi-lender pools subject to a de minimis limit, like jumbo loans.
Given that these high LTV cash-out refinances don’t appear to prepay too differently than other refinances it seems that exclusion from the multi-lender program is probably overkill. It is also unusual that these loans would be excluded while, for example, modified loans would continue to be permitted in multi-lender pools without limitation and jumbo loans would be subject to the de minimis limit.
Instead, imposing a de minimis limit might be sufficient. Although aggregate production of these loans is well below 10%, it is important to remember that the de minimis threshold is applied per-lender within each pool. So any abusive lender that originates far too many of these loans would find much of their supply locked out of the multi-lender pools, while responsible lenders with more reasonable volume would still be able to deliver their production into the multi-lender pool.