A Loss Severity Model for Residential Mortgages

Abstract

The loss severity model, combined with prepayment and default models, can be used to project loss-adjusted cash flows for mortgage pools and deals. In addition, the loss model can be used for identifying potentially risky loans in a portfolio of subprime loans. In this article we analyze loss severities on mortgage loans with varying borrower and loan characteristics, and we develop a predictive model for projecting losses. We also show that simplistic models that rely heavily on loan-to-value ratio produce unreliable loss severity projections and that a more comprehensive approach, such as the one described in this article, is required for accurately estimating credit risk.

TOPICS: MBS and residential mortgage loans, big data/machine learning, legal and regulatory issues for structured finance