The days of relying on static, off-the-shelf prepayment models and taking a credit rating agency’s rating of a mortgage-backed security at face value are long gone. While the underlying mortgages in new agency securitizations are some of the cleanest, most well underwritten loans that investors have seen in a generation, Wall Street isn’t taking any chances.
Investors, intent on reclaiming their destinies, are demanding new ways to evaluate both existing and new MBS deals. That higher level of scrutiny requires extensive data, robust analytics and a multitude of models that can be adjusted in near real-time to anticipate how internal and external forces will impact an investor’s book.
All risk isn’t the same, particularly between agency and private-label securities. “With agency MBS the investor is taking mostly interest rate related prepayment risk,” while “nonagency investors are taking primarily home price related credit risk,” said V.S. Srinivasan, Barclays managing director of securitized products modeling
To help Barclay’s trading desk and its clients better analyze broad market trends and their own portfolios, the firm developed a technology called Barclays Live. It’s a Web-based platform that includes predictive models and other analytics tools, as well as research and data indices across every major asset class and investment, including mortgage securities.
“The trick of good analytics is to have a good term-structure model, which calibrates every night, and a good prepayment model, which produces realistic cash flows,” said Srinivasan.
The term-structure model forecasts changes in interest rates. Combined with prepayment models, investors are presented with a variety of scenarios to help them predict how their investment will perform.
But developing these models is no longer a matter of simply tracking changes in mortgage interest rates, like it was when borrowers could more easily refinance.
Now, a borrower may have a lot of incentive to refinance, but weak credit scores, underwater existing mortgages with high loan-to-value ratios and other credit conditions are keeping borrowers from refinancing, leaving many traditional models inaccurate.
“You have to calibrate your models to what has happened over the last two or three years,” Srinivasan said. “You want to look at the collateral characteristics of a pool or group of loans and ask what percentage of these borrowers can get a mortgage. And conditional on them being able to get a mortgage, what percentage will actually refinance?”
