Property Valuation Data from the GSEs
Among the relief measures rolled out by housing regulators in responding to the coronavirus pandemic was a March 23rd directive from FHFA to the GSEs to “leverage appraisal alternatives” in underwriting loans given the conflict between social distancing and the interior inspection appraisals often require. By sheer coincidence, this instruction more or less coincided with the March release of loan-level data on the different property valuation methods used for loans issued in GSE MBS. This data set allows us to track the relative share of different property valuation methods and to also explore their link to borrower prepayment behavior. Table 1 gives an overview of the share of these different methods for loans in 30-year Fannie Mae MBS (the numbers for Freddie Mac are similar). Without getting into the technical definitions of the different methods, it’s clear from the table that the overwhelming majority of properties either require an appraisal or receive an appraisal waiver.
Unpacking the aggregate data by loan purpose shows that appraisal waivers are much more prevalent for refinancings versus purchase mortgages which is not surprising given that the bar for not appraising a property backed by a purchase mortgage is a lot higher. Also, as time has gone by, waivers have become more popular. The figure below shows that the purchase share of appraisal waivers is around 5% for both GSEs and has averaged 25-30% over the past year for refinance loans, noticeably higher than the 5-10% seen in 2018 shortly after the waiver program was first introduced. We should expect the share of appraisal waivers to trend higher in the coming months.
Who Gets An Appraisal Waiver?
What types of borrowers get an appraisal waiver? Table 2 summarizes some collateral attributes of mortgages that relied on an appraisal versus those that received a waiver. The focus is on refi mortgages for clarity. Generally speaking, mortgages that qualified for appraisal waivers were made out to borrowers with higher loan balances and FICO scores, and lower debt-to-income ratios, in short borrowers with a better credit profile. One item in the table deserves special attention, the SATO (spread-at-origination) measure: this serves as an approximation of how much above-market a borrower's rate is versus the average. Essentially, the SATO encapsulates a lender's assessment of the credit risk posed by a particular mortgage -- higher SATOs correspond to loans with a greater likelihood of default.
Comparing Prepayment Rates on Appraisals and Waivers
As a first pass towards understanding the prepayment characteristics of loans that utilized different property valuation methods, we fix all loans that were pooled in a certain coupon and vintage year and then separate them into two groups based on which property valuation method they utilized. The results are summarized in the figures below and show that loans that have received an appraisal waiver prepay faster by about 10-15% CPR than those that don't.
The estimated prepayment differential is potentially not very sharp because, as we saw above, the comparison between loans receiving an appraisal versus those that don't is confounded by the fact that the two different collections of loans may have different loan sizes, credit scores, and incentives. We develop a sharper estimate by plotting the prepayment rate as a function of incentive (the "S-curve") for the two different methods while also controlling for servicer, credit scores, OLTV ranges, loan purpose and loan size. These S-curves suggest that even after controlling for a number of salient characteristics, the difference still averages about 10% CPR.
Even though this analysis is driven more by slicing-and-dicing data rather than modeling, it is quite suggestive. Still, it's not immediately clear why there's such a big difference in the prepayment rates of the two different groups of loans. On the S-curves above, 10% CPR is equivalent to about a 25bps change in incentive but saving the cost of an appraisal is probably only worth about 5bps.
One key lies in the table of collateral characteristics which showed that on aggregate the SATO on loans that received an appraisal was about 29bps higher than those that received a waiver. What the SATO is reflecting is that borrowers who have appraisal waivers typically qualify for better mortgage rates, rates that are on average 29bps lower (across all collateral characteristics). In other words, using the same mortgage rate to calculate incentive for both collections of loans is understating the incentive for waivers and overstating it for appraisals.
To normalize for this fact and assuming that these rate differential continued to be available we redo the S-Curves using the SATO-adjusted incentive (defined here as Loan Rate - (Mortgage Rate + SATO)). The difference between the two S-curves compresses significantly for Quicken-serviced loans but there's still a noticeable residual.
We made the strong assumption that the SATO differential at origination remained constant throughout. In fact, as the graph below shows, the differential expanded by about 10bps in 2019, thus potentially explaining the full gap between the observed prepayment rates.