Self-Reporting Under SEC Reg AB and Transparency in Securitization : Evidence from Loan-Level Disclosure of Risk Factors in RMBS Deals

This paper is designed to illustrate the limitations and potential bias in securitized residential mortgage data and examine the importance of such data issues for typical studies of RMBS market and the financial crisis. We use trustee data on mortgage characteristics provided by BlackBox Logic – the BBx data – to study the extent to which undisclosed mortgage characteristics distort the available data and impact risk analysis of RMBS collateral pools. We illustrate that substantial amounts of loan characteristic data in crucial fields like Occupancy, Property Type, Loan Purpose, and FICO are missing from the trustee data. The frequency of missing values is staggering, ranging from just under 9% for Property Type to 29% for FICO, up to almost 85% for Originator Name, all variables used in recent studies. The omissions are correlated to some degree with the securitization sponsor and even more dramatically with the identity of the deal trustee. Analysis of RMBS collateral should be built not upon the entirety of mortgage databases, but on stratified samples and should otherwise control for important sponsor and trustee fixed effects. The revisions for Regulation AB which require loan-level disclosure should be adopted in order to standardize mortgage disclosure. This is the first paper that examines selection bias in loan characteristics relied upon for a wide variety of mortgage market research that has substantially affected policy decisions in the post-crisis era