As traditional prepayment models, which assume a generic borrower base, become unfashionable, Bear Stearns has launched a new model focusing specifically on loan or borrower attributes.
Standard models have generally looked at distinctions in loan type, loan seasoning, and the average response of the borrower to refinancing opportunities - features generalized across borrowers in the agency market. Bear's new model considers other factors, such as home price appreciation, loan size and rate premium for analyzing the agency sector.
"One of the central tenets of our new agency model is to, wherever possible, explain prepayment behavior using actual loan attributes (calculated at the pool level) rather than identifying other variables that may proxy this behavior," wrote analysts from the firm in a report.
Bear added that the approach allows them to measure many of the anecdotal themes that dominate "intuitive" prepayment analysis - such as fast/slow servicers, the influence of refinancing efficiency gains, cash-out refinancing, and geographic influences - through measurable changes in these loan attributes.
For instance, in terms of fast/slow servicers, traditionally, fast servicers tend to originate higher balance loans and higher WAC loans. So the important factors in this instance are not the solicitation efforts being made by the servicer, but rather the loan size and the weighted average coupon. That one servicer tends to be "faster" than another could therefore be explained and/or quantified through these loan attributes.
Loan size
"One of the things we want to address here is as the pool composition changes within the mortgage market, we should have a model that stays current with those changes," said Dale Westhoff, senior managing director at Bear. "For example, one of the major recent changes is that pools are being originated today with much higher average loan balances."
In the last few years, the loan sizes of pools have gone up dramatically as home prices have increased and as the agencies have increased their limits. Average loan sizes in agency pools used to be in the $100,000 to $125,000 range. Currently, average loan sizes have gone up to as much as $160,000 to $180,000. This will have a significant effect on a pool's prepayment characteristics because larger loan balance pools tend to be more negatively convex.
Home price appreciation has also been factored into the new model. Bear has calculated a home price index for each pool (which is based on state-level breakdowns of the pools provided by the agencies and home price indices for each state) and will be updating the index every month as home prices increase. The index provides a good measure of how much equity the borrowers backing a particular pool have in their property. To the extent that home prices increase, the model is going to forecast a greater component of cash-out refinancing.
Cash-out refinancings have been a major factor in prepayment behavior over the last few years. Borrowers have been cashing out the equity in their homes, which increases as home prices go up, to pay off and consolidate other higher-interest, non-deductible debt, such as credit card and student loans. Even if a mortgage rate rises in a cash-out, it might be lower than the consolidated rate of their other debt.
The new model is also very sensitive to pools that have credit-impaired or Alt-A borrowers, which are increasingly showing up in Agency pools. The model measures this factor by looking at the rate premium that these borrowers paid over the conforming mortgage rate at the time the loan was originated. To the extent that the borrowers paid 50 or 100 basis points over the conforming rate, this means they probably fall into the non-standard category, thus are most likely going to demonstrate less sensitivity to interest rates.
Aside from these factors, the model also accounts for the current popularity of the hybrid mortgage product as a refinancing alternative, as well as the number of mortgages which are actually refinance loans versus a pool's purchase loan concentration.
Although analysts do not know the actual collateral characteristics of the mortgages being delivered in a TBA pool, to value the TBA market, analysts could look to the latest allocation information. This data shows what is being delivered into the TBA market in a particular week. Analysts use these average numbers to value a current security where there is no collateral information available. This keeps the model current on the underwriting guidelines that the agencies employ and the pool compositions, especially as the pools become more heterogeneous.
