Mortgage Appraisal Quality, Risk Layering, and Loan Performance

Abstract

Poor-quality mortgage appraisals are thought to have played a role in real estate market downturns from the 1980s to the 2000s. Prior research measuring appraisal quality based upon ex post audits of individual appraisals, however, have been limited to small samples of limited power and flexibility. The present paper uses digital distribution theories to derive appraisal quality indicators for 6.5 million mortgages originated from 2000 to 2007. We establish that appraisal risk is multi-layered, so that two appraisal quality indicators are more informative than one. Conditional mixed process models of default assuming endogeneity in appraisal quality choice show that poor quality appraisals typically raise default risk by about 1% (of a total of 4% - 10%) in non-crisis years, increasing to 2%-3% (of a total of roughly 20%) in 2006-2007. Interacting appraisal quality indicators with payment-reduction affordability features to account for risk layering, however, results in marginal effects up to 9% (of a total of roughly 30%) in 2006-2007. Moreover, while adverse appraisals increase default risk in all years, affordability features either have no effect upon or reduce default risk outside the crisis, suggesting that affordability features may benefit borrowers when accompanied by careful underwriting.

Keywords: appraisal, Benford’s Law, cluster analysis, cognitive reference theory, financial crisis, machine learning, mortgage default, numeric distribution theory, risk layering

JEL Classification: C55, D12, E44