Fitch Ratings' RMBS group introduced its new RMBS credit enhancement model last week, which facilitates regression analysis on three different sectors: prime, Alt-A and subprime pools.
Through this new model, analysts are able to re-estimate the probability of default on new origination and performance data, look at "risk premium" rather than the coupon of the loan (assessing the relative cost and not the absolute cost of the mortgage), make a deeper analysis of FICO scores (specifically for subprime loans) and consider the timing pattern of defaults in terms of the sectors they occur in.
Fitch differentiates between the three different sectors based on the data at the point of origination on loans made within a 36-month time frame. In the model, a bad loan is characterized as a loan that is 90 days late or worse within varying performance windows for each of the three sectors. For subprime and Alt-A, Fitch gives a window of three years of origination. In terms of prime mortgages, Fitch includes a window of four years of origination.
Fitch said that "bad" loans tend to pay off suddenly, such as when the mortgage is refinanced out of delinquency or when it is sold for a small loss. "Abrupt resolutions are highly correlated to performance within the windows," analysts said. For example, most bad mortgages abruptly resolve 90% of the time in subprime, 80% in Alt-A and 70% in prime.
The model also made use of LTVs and FICO scores to differentiate between the sectors. In pools evaluated over the past year, Fitch has seen that prime pools are becoming "more prime" as FICO scores rise and LTVs decrease. However, Alt-A pools are becoming marginally riskier as FICO scores decline and LTVs rise, although the makeup of the pools tend to be more volatile. In terms of subprime, the average LTV has moved marginally from 75 to 79.
Comparisons within the sectors using FICO and LTV data also reveal that while particular pools may have similar weighted averages, the distributions could be different. For example, comparing the Prime 1 and Prime 2 pools, 25% of the borrowers within the Prime 1 pool had an LTV that was less than or equal to 60 and a FICO score greater than or equal to 720, which is the lowest risk category. On the other hand, in the Prime 2 pool (which is still made up of very high-quality borrowers), only 19% of the pool is in the lowest risk category.
An important component of the model is looking at the risk premium as the relative interest cost of the loan. In determining which sector a borrower belongs to, one has to look at the prevailing interest rate at the time the loan was funded or when that rate for the mortgage was locked in. For instance, if the prevailing market interest rate is 7.75% and the borrower has an 8% coupon, the lender would consider the borrower good because the relative cost is only 25 basis points. However, if prevailing interest rates at the time of origination were 5% and the borrower has an 8% mortgage, the lender will view the borrower as a much greater risk because the relative cost in that
case is 3%.
In terms of loss severity, Fitch has developed models together with Economy.com. One interesting component of the new model is that it allows analysts to consider both borrower-paid and lender-paid insurance without double counting.