[1]New York-based credit rater Kroll Bond Rating Agency (KBRA [2]) has updated its residential mortgage default and loss model, incorporating a new methodology that projects loan-by-loan default, loss, and prepayment on residential loans in order to track non-agency residential mortgage-backed securities (RMBS), KBRA announced [3].
The new methodology uses revisions that reflect additional data analysis and evolving origination trends, and is an update to KBRA's RMBS model methodology originally released three years ago in January 2012.
Substantive revisions were made to the model in three areas:
- Reduced default expectations for purchase loans. KBRA adjusted the model's treatment of purchase loans "to reflect lower default expectations relative to equivalent refinancing mortgages," according to the methodology report. This adjustment was made based on KBRA's analysis of both jumbo and conforming mortgages which found that the traditional benefits of purchase loans remain well established.
- Revised timeline and expenses for liquidated loans. To better reflect historical data, the new model reduces the amount of time a loan is assumed to be REO. While this adjustment will result in a reduction of carrying expenses, "it also has the effect of better aligning peak liquidation periods with the peak home price stress, therefore, the net effect is to increase loss severity," KBRA wrote in the methodology report.
- Penalty for high debt-to-income ratio loans. KBRA wrote that although the model does not contain specific DTI-based risk parameters, they believe high DTI loans can "bear significant incremental risk," and therefore the agency assigned an additional default penalty to loans with back-end DTIs that exceeded 45 percent.
The original model continues to serve as the basis for KBRA's PD (probability of default) model. The updated model is a transition model that uses CoreLogic's LoanPerformance (LP) database as the basis for a regression analysis of more than 20 million residential mortgage loans and their performance from 2000 to 2010. The revisions made to the mortgage default probabilities in the latest methodology report have been implemented as adjustments to KBRA's original recession model.
The methodology report [4], entitled the Residential Mortgage Default and Loss Model, includes data such as calculation for loss severity, expected loss for defaulted loans, rationale and definition for levels of default, and loss stress associated with each rating category. The report documents the elements as well as the predictive power of the default and prepayment model.
The report also discusses additional modeling considerations such as geographic concentration and cash flow structure analysis.