Conventional speeds jump, Ginnie Mae may remain fast
admin | October 11, 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.
Interest rates since mid-September have dropped back to levels reached in August and have kept investors worried about elevated prepayment speeds. Conventional speeds in September increased overall by more than 10%, primarily driven by pickups in newer vintage cohorts and especially by 2018 production. Ginnie Mae speeds lagged but look likely to rise in the months ahead. Combined with changes to bank liquidity regulations approved by the Fed on October 10, faster speeds could hurt Ginnie Mae performance.
Overall Fannie Mae 30-year speeds increased 11.9% to 19.3 CPR from 17.4 CPR and Freddie Mac 30-year speeds increased 14.3% to 20.1 CPR from 17.8 CPR, both somewhat faster than expectations. Although driving rates were lower in September, seasonal turnover begins to slow and there were two fewer business days compared to August. Both factors suppressed some of the rate-driven pickup in prepayments. Loans are most refinanceable between six and 24 months of seasoning, and the surprisingly fast 2018 vintage is almost entirely in this seasoning band.
Exhibit 1: Conventional speeds jump in September
Ginnie Mae speeds were more muted, partially due to better call protection at low WALAs. However the MBA Government Refinance Index remains elevated and the gap over the conventional refinance index has widened. This suggests that Ginnie Mae speeds will remain elevated and that 2019 vintage Ginnie Mae speeds are poised to increase as the loans season past six months.
Ginnie Mae II prepayment speeds increased only 2.3% to 23.0 CPR from 22.5 CPR. Many 2019 pools are still not seasoned enough to refinance, and many of the seasoned cohorts are already prepaying extremely quickly. The government refinance index picked up earlier than the conventional refinance index, peaked higher, and has not fallen as much as the conventional index over the last month.
Prepayment speeds should pick up by about 10% in October primarily due to 2.5 additional business days. Interest rates increase roughly 8 bp on a 30-day lag but fall roughly 7 bp on a 45-day lag, which should boost speeds on more seasoned cohorts. Speeds should slow by about 10% to 15% in November, which is a very short month due to the holidays, although lower lagged interest rates will support speeds on newer vintages. At current interest rates roughly 50% of agency mortgages are refinanceable (at least 75 bp in-the-money).
The 2018 conventional MBS vintage surges
Conventional 30 year pools issued in 2018 prepaid much faster in September. The S-curve is steepest for loans 6 to 24 months seasoned (Exhibit 2), and 99% of the 2018 vintage is within that seasoning band.
Exhibit 2: The S-curve peaks between 6 to 12 months seasoning
However, 2018 speeds were even faster than the S-curves suggest. At the steepest part of the S-curve a 20 basis point rally (mortgage rates fell roughly 20 bp in September on a 30-day lag) suggests speeds should increase roughly 4.0 CPR. There were two fewer business days, which should have lowered speeds roughly 1.5 CPR, which brings the expected speed increase to about 2.5 CPR.
The 3.5%s and 4.0%s of 2018 both blew past this estimate, increasing 5.8 CPR and 7.1 CPR respectively. Interest rates have reached recent lows and this is the first refinance opportunity many of these borrowers have faced. So it is likely that the S-curve is steepened due to a media effect and increased solicitation efforts.
The 2017 vintage was less responsive than the 2018 vintage
The 2017 vintage 3.5%s and 4.0%s are prepaying slower than the respective 2018 vintage cohorts—the pickup from August to September was smaller and the absolute levels are lower. This is primarily due to seasoning—the 2017 vintage is almost entirely over 20 months seasoned, while the 2018 vintage is mostly less than 2 years seasoned.
There are reasons other than seasoning that could explain slower prepayments, but do not hold up to scrutiny. One possibility is that the 2017 vintage could be slower from burnout. A loan is “burned out” when the borrower has passed up opportunities to refinance. Those loans prove to be less likely to refinance the next time interest rates fall. However the 2017 vintage never had an opportunity to refinance prior to 2019. In fact the 2017 vintage has averaged only 6.8 CPR lifetime through March 2019, which is not much above baseline turnover levels.
The late 2018 vintage pools also have a much higher gross WAC than previous vintages, which also contributes to faster speeds. However, comparing the two vintages using S-curves controls for the gross WAC differences and shows that the 2018 vintage had a much steeper S-curve (Exhibit 3).
Exhibit 3: The 2018 vintage S-curve was much steeper than the 2017 S-curve
Another reason the 2017 vintage might be slower is spread at origination (SATO), which measures how much higher or lower a loan’s note rate is compared to other loans originated at the same time. A higher SATO indicates a loan was charged higher fees by the GSEs and/or lender, and is likely a worse credit. Higher SATO loans will prepay slower initially (if the borrower’s credit cures then speeds could pickup). A 2017 loan will have a higher SATO than a 2018 loan with the same note rate since rates were lower in 2017. However, excluding loans with SATO>25 bp from the S-curves shows that the speed difference remains (Exhibit 4).
Exhibit 4: The 2018 vintage S-curve was steeper after removing high SATO loans
This leaves seasoning as the most likely reason that the 2018 vintage prepaid faster than the 2017 vintage. Loans that were originated more recently are easier to refinance—documentation is more readily available and items like appraisals might still be usable. These borrowers are more likely to be aware of mortgage rates and more likely to get a call from their loan officer. Once a loan is a couple of years of seasoned it is less refinanceable even if it hasn’t yet had a refinance opportunity.
Investors should look at speeds from loans between 6 and 24 months seasoning for guidance regarding peak prepayment speeds. Prepayments can increase to peak levels very soon after a loan surpasses 6 months seasoning, but speeds slow quickly after 2 years of seasoning. The S-curve will flatten even if a loan has not had the opportunity to refinance, suggesting that a portion of the prepayment response is “lost forever” if a loan survives past 24 months seasoning without experiencing low interest rates.