Speeds remain surprisingly fast in May
admin | June 5, 2020
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.
Agency prepayment speeds remained extremely fast in May, again exceeding expectations. Many market participants had anticipated slower month-over-month speeds. The MBA Refinance Index had fallen considerably since mid-March, and May had one less business day than April. The recent sudden increase in delinquencies and loans entering forbearance also weighed on speed expectations. It appears many borrowers that applied to refinance in March faced very long processing times and eventually closed in May. This boost to prepayment speeds should dissipate in June, but speeds are likely to remain elevated for at least several months.
Fannie Mae 30-years speeds only slowed 2.3% and Freddie Mac speeds came in flat in May. Ginnie Mae II speeds increased 4% overall. Across all three programs, seasoned vintages tended to come in faster. Because seasoned pools generally have higher coupons, speeds in higher coupons showed larger month-over-month increases.
Typically, it is more work to refinance a seasoned loan. Very new loans may be able to reuse some documentation and may be more likely to qualify for a property inspection waiver. Seasoned loans also tend to be smaller in size and less profitable to refinance. Loan officers typically put higher priority on refinancing larger loans, so, when facing capacity constraints, closing times will extend for smaller loans. A loan officer might even assist borrowers by paying rate lock extension fees, and it would make more sense to pay those fees on the smallest possible loans.
Exhibit 1: Speeds remained near peak levels in May
March mortgage applications boosted May prepayment speeds
The MBA refinance index has steadily fallen since mid-April and is well below the peak levels from early March (Exhibit 2). The falling index, one less business day, and the increase in delinquent loans and loans on forbearance plans pointed towards 20% slower speeds in May. But if some loans took a very long time to close, 60 days or even longer, they would become May prepayments. A simple analysis can quantify this boost to May prepayments.
Begin by assuming that the refi index should affect speeds in the following month. For example, the change in the refinance index from February to March would correspond to the speed increase from March to April. In this case the refinance index increased 66%, which implies refinance speeds should increase roughly 50% assuming 75% of loans eventually close. But prepayment speeds only increased 30% in April, leaving a 20% gap. Some of this difference might be explained by slower turnover but likely a significant amount of those refinances closed in May. These additional prepayments offset the slowdown implied by the refinance index using typical 30-day and 45-day lags, and explain why speeds were flat instead of significantly slower.
Exhibit 2: MBA Refinance Index
This has implications for prepayment speeds going forward, which should remain roughly unchanged for the next couple of months assuming steady mortgage rates. It appears the effects of elevated March refinance activity were fully accounted for in the April and May prints, so the boost to prepayment speeds should subside. On the other hand a few factors point towards faster prepayment speeds in June—mortgage rates are lower on 30-day and 45-day lags, lower refinance volume should give originators more capacity and put pressure on primary-to-secondary spreads, purchase activity is recovering, and there are two additional business days in June. The net effect should be unchanged prepayment speeds over the next couple of months.
There are a couple of other factors that could lead to slower prepayment speeds, but given the surprises of the last couple of months it seems unwise to assume that will happen. The first would be that possibly the accumulation of delinquent loans and loans in forbearance plans will finally pull prepayment speeds lower. The second is that the MBA refinance index has recently been low compared to the level of mortgage rates (Exhibit 3).
Exhibit 3: MBA Refinance Index vs. 30 Year Mortgage Rate
The red triangles show that the refinance index was extremely high compared to reported mortgage rates in March, which is a big reason the April and May prints came in much faster than expectations. The relationship appears to have returned to normal in April, but in May swung the other direction. If prepayments continue to track the refinance index the market could finally receive a pleasant surprise of slower than expected speeds.
A word on the GSEs’ forbearance and delinquency disclosures
This month the GSEs began to disclose information about borrower assistance plans for all active loans in pools. Most interesting is the number of loans in forbearance. Fannie Mae’s data is being reported on a one-month lag and shows 6.0% of 30-year loans in forbearance. This is consistent with the MBA’s forbearance survey from April 27 to May 3, which showed 6.08% of conventional loans were in forbearance. Freddie Mac’s disclosure is not supposed to be lagged—their announcement made no mention of a lag—yet they reported merely 3.5% of their loans in forbearance plans. That is far below the combined 30-day and 60-day delinquency rates, which are 5.62% for Fannie Mae and 5.30% for Freddie Mac. Something seems amiss with the Freddie Mac data.
The delinquency numbers have moved much closer to the forbearance numbers. This was expected. Most loans went into forbearance plans in April, likely after making the payment due April 1. These loans will immediately be reported as being in forbearance but will not be identified as 30 days delinquent until June 1, 30 days after the due date of the missed payment. The May report’s 30-day delinquency mostly reflected loans that went on forbearance in March, and the June combined 30-day and 60-day numbers reflects forbearance begun in March and April. Since the pace of new loans being placed on forbearance plans has slowed considerably it is likely that the delinquency and forbearance numbers will trend even closer in the July report.
A deep dive into the June delinquency data
The following charts examine the GSEs’ pool-level delinquency data for various collateral buckets. The data confirms various trends highlighted last month: delinquency rates increased more for loans with low FICO scores, high LTVs, higher SATO, larger loan sizes and higher coupon:
Exhibit 4: More Fannie Mae loans became delinquent in May
Exhibit 5: The disparity between low and high FICO pools grew in May
Exhibit 6: The increase in DQ is similar across most of the LTV buckets. The rate of increase is also slower than last month. High LTV pools continue to have higher delinquency rates than low LTV pools and the difference between them remained relatively similar in June. Overall delinquencies increased, but slower than in April
Exhibit 7: High SATO pools had larger increases in DQs than low SATO pools
Exhibit 8: Pools backed by New York collateral had larger increases in DQs.
Exhibit 9: Investor pools had a large increase in DQs again in May.
Exhibit 10: High coupon pools generally contain borrowers with a higher payment and often lower credit characteristics. These pools also had a larger increase on DQs than lower coupon pools.
Exhibit 11: Larger loans continued to become delinquent more quickly than lower balance pools. One explanation is that unemployment benefits and the additional benefits of the CARES Act will have a relatively larger effect on lower income borrowers.
Exhibit 12: Wells Fargo once again had the highest DQ rate as well as the largest increase of originators and servicers. The DQ rate on some other lenders, like Provident, are beginning to catch up. This is likely due to the timing of when forbearance began and how efficient the servicer was at stopping automatic payments.
Thanks to the APS Data and Analytics team for their support on the delinquency analysis.