Latest Developments in Mortgage Analytics
for Investors

Abstract

Raj Dosaj explained various factors that contribute to the accuracy of home price indexes, such as geo-granularity, attributes of the McDash database, and the Lender Processing Services (LPS) Credit Bureau Module. Michael Bradley explained a set of disposition models developed by CoreLogic that estimate distressed sale values at various stages of the liquidation process that he believes are more accurate, less expensive, and can be obtained more quickly than broker price opinion (BPO) valuations. Dan Abraham discussed the use of Lewtan’s address-specific rental automated valuation models (AVMs) and how those models are reflected in statistics at the neighborhood and ZIP Code levels. Ron D’Vari discussed challenges with loan-level models for U.S. RMBS (residential mortgage-backed securities), including too many parameters and not enough time to validate interdependencies and the dominance of macro factors, such as interest rates and HPIs (house price indexes), and how risk-based cohorting provides a solution to incorporate loan-level performance. Brian Grow explained Morningstar’s approach to rating structured securities, in which issuers pay for initial ratings, investors pay for surveillance, a proprietary loan-level credit model is used, and all rated securities are re-rated monthly.