By the Numbers
Finding MBS with faster prepayment speeds below par
Brian Landy, CFA | March 4, 2022
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.
Higher mortgage rates this year have understandably spun up a lot of new interest in MBS with extension protection. More than 60% of conventional 30-year MBS now trade below par, a sharp increase from 4% at the start of the year. Investors need to find collateral that prepays faster when borrowers cannot refinance into a lower mortgage rate. The most interesting candidates include loans from specific states, loans with lower balances and loans with lower FICO scores. Faster speeds could add as much as 16/32s of value, and that could go higher if rates continue to rise. And combining attributes may boost pay-ups further.
Typically, there is limited information on prepayment speeds in MBS trading below par since mortgage rates have generally trended lower for the last 40 years. But refinance activity in 2020 and 2021 created a modest cohort of 30-year 1.5% pools that have generally traded at or below par since origination. The 1.5% pools provide some insights. In studying various collateral attributes, it is important to control for the impact of other attributes. For example, estimating the effect of state on prepayment speeds should control for loan size, FICO, servicer, and other attributes. Otherwise, the impact attributed to state might only reflect the influence of loan size. The effect of loan size needs to control for state and all other attributes, and so on.
The impact of geography
State has a large influence on discount prepayment speeds. Loans from Arizona, all else equal, prepaid roughly 25% faster over the last two years, for example (Exhibit 1). On the other side, discount loans from New York prepaid more than 30% slower. Housing turnover is a primary contributor, as many of the states with faster prepayment speeds have hotter housing markets and net inflows of residents. Cashout refinances likely also play a role, especially in states with high home price appreciation. Appreciation can drive housing turnover, but states with tight housing supply may not reap the full benefit. But cashout refinance activity is not constrained by supply and have a tighter correlation with HPA.
Exhibit 1. Discount prepayment speeds vary across states
Loans from Arizona, prepaying roughly 25% faster than the average loan, should be worth 16/32s more in a FNCL 1.5% pool (Exhibit 2). This is determined by calculating the FNCL 1.5% TBA OAS using Yield Book’s prepayment model, then dialing the model with 25% faster discount prepayment speeds and calculating a price using the TBA’s OAS. The dialed pool is projected to prepay 6.7 CPR for life, compared to 5.6 CPR for the TBA. That is a small absolute CPR difference and would not be very meaningful in a fast prepay refinance environment. But small speeds difference matter in a turnover environment and especially in a security now trading below $93-00.
Exhibit 2. A 25% boost to turnover speeds adds 16/32s to a FNCL 1.5% pool
The theoretical pay-up depends on the coupon and amount of prepayment difference (Exhibit 3). The pay-up is bigger for pools that trade farther below par, while faster turnover has negligible effect on FNCL 3.0%s that are currently trading slightly above par.
Exhibit 3. Price sensitivity varies with coupon
The impact of credit score
Borrowers with lower credit scores also tend to prepay faster in a discount environment. These borrowers may be new borrowers that can benefit from credit curing to move into a larger home. They also may be more likely to access home equity using a cashout refinance. And default rates may also be higher and contribute to faster discount speeds. Borrowers with a FICO score of 700 or lower prepay from roughly 7% to nearly 15% faster than borrowers with higher scores (Exhibit 4). As indicated earlier, a 5% faster speed in a 30-year 1.5% pool currently is worth an estimated 3.5/32s more than TBA, a 10% speed is worth 6.5/32s and a 15% speed is worth 9.5/32s (Exhibit 3).
Exhibit 4. Low FICO loans prepaid faster out of the money
The impact of loan size
Smaller loans also tend to prepay faster in a discount environment, while larger loans prepay slower. Loans of $85,000 or below prepay roughly 10% to 15% faster than average while loans of $500,000 or above prepay between 7% go more than 10% faster. These borrowers with smaller loans often are in starter homes and will eventually need to buy a bigger house, driving faster prepayment speeds.
Exhibit 5. Low balance loans have faster discount prepayment speeds
The impact of LTV
Discount prepayment speeds tend to be faster for loans that have current LTV’s below 80% but still are paying mortgage insurance (Exhibit 6). The borrower has additional incentive to refinance to stop paying insurance premiums. Although Fannie Mae and Freddie Mac allow cancelling mortgage insurance in some instances, lenders are not allowed to solicit borrowers. It is possible that borrowers refinance because they are unaware they might be able to cancel their mortgage insurance.
Exhibit 6. Loans with MI prepay to cancel mortgage insurance
The impact of servicer
Some servicers have a large effect on refinance speeds, and this might carry over to discount prepayment speeds (Exhibit 7). This shows the residual effect of some of the largest conventional servicers on discount speeds over the last two years. Quicken and loanDepot tend to prepay much faster than other servicers, while Chase is lagging with much slower speeds. However, it is still uncertain whether the faster servicers will be able to maintain that behavior.
Exhibit 7. Servicer may have a large influence on discount speeds
The impact of loan purpose
Rate/term refinances prepay slightly faster when out-of-the-money (Exhibit 8). Refinance loans have additional seasoning that isn’t reflected in the loan age when compared to purchase loans. Purchase loans and cashout refinances tend to prepay a little slower. The cashout result is surprising since these loans also have additional seasoning that isn’t reflected in the loan age. It’s possible that the limited amount of data is insufficient to properly identify this effect.
Exhibit 8. Rate/term refis tend to have faster discount speeds than other loan purposes
The MBS market is just starting to figure out the nuances of prepayments in pools trading below par where borrowers have no rate incentives to refinance. Those pools also happen to be ones where, except for loan balance and some geographies, price premiums to par are low or nonexistant. Of course, it may be difficult to find pure plays on attributes such as geography in pools trading in secondary at a discount. The best opportunities are likely to be in LTV or FICO pools, where investors can find pure plays and where price premiums to TBA are small.