Nearly a half decade after the start of the mortgage crisis, portions of borrower vetting for securitized mortgages are still being done primarily by loan originators. This causes a potential risk management gap, particularly for non-Fannie Mae and Freddie Mac-backed mortgages that are bundled and sold as mortgage-backed securities.
“The bond market needs to focus more on the front-end origination of mortgages. For certain, they have to do the same type of due diligence review as lenders. It can be an automated review to look at a whole portfolio to see which loans were good and which were bad. I don’t see that tech on the buy side," said Christine Pratt, a senior analyst at Aite Group who specializes in lending and credit risk.
The borrower vetting gap exists more for nonagency mortgages bundled into mortgage-backed securities and sold into the secondary market. Nonagency mortgages refer to loans that are not backed by government sponsored enterprises such as Fannie Mae and Freddie Mac, and thus do not fall under the agencies’ underwriting and credit risk criteria, which has tightened in the wake of the crisis.
The nonagency mortgage market, which has traditionally referred to subprime loans, alt-A loans and jumbo loans, has been relatively stagnant for the past few years, with yearly issuance of nonagency mortgage-backed securities falling from a high of $1.2 trillion in 2005 to less than $100 billion in early issuance following the 2008 collapse. But as the residential mortgage market starts to recover, there are some signs of a resurgence, and Pratt says there need to be more “checks and balances.”
After the crisis, there was an expansion of credit risk tech that was designed to speed mortgage processing while performing better due diligence on borrowers. This new tech included broader information such as regional price changes, behavioral tendencies to gauge a borrower’s propensity to pay, automated employment and income verification, and even some early use of social network analysis to get a sense of a borrower’s overall financial health.
Mortgage lenders and buyers are normally reluctant to talk specifics about credit risk, given the sensitivity of the issue. Two of the largest providers of mortgage risk technology, LPS Applied Analytics and CoreLogic, both say the tech being used by the buy side and sell side are different.
LPS Applied Analytics’ buy-side tech leverages models that focus on gauging prepayment risk, and also determine the potential impact on a pool of mortgages resulting from the evolution of government mortgage programs, refinancing market changes and event risk such as pressures on MBS pools resulting from the European debt crisis. Raj Dosaj, a vice president at LPS Applied Analytics, says its sell side tech collects data from servicers and “they don’t require all of the information that originators have. It’s not a nefarious plan. There’s no incentive to change how it’s being done currently.” Dosaj says there’s been some policy talk about increasing disclosure thresholds for originators.
“Origination guidelines are tighter now, but there is still a murky world to understand now things are being processed. It’s not transparent. We have large data servers and can glean what’s going on, but the detail of why people are rejected for loans and what the pricing is, is not fully transparent,” Dosaj says. Dosaj stresses that for agency mortgage-backed securities, Fannie Mae and Freddie Mac are taking on all of the credit risk, so the “thing you focus on as an investor is prepayment risk, so having that information is helpful for prepayment, but it’s not the same risk as when you are exposed to credit [risk].”
CoreLogic’s suite includes risk modeling for price changes and rate changes, and focuses on loan-to-value ratios. The firm is interested in ensuring accurate appraisals and spotting appraisal fraud, so that the stated value of loans that are being originated and subsequently sold is accurate. “If the appraisal is wrong, the LTV is wrong, and all bets are off,” says Michael Bradley, senior vice president of modeling and analytics for CoreLogic.