ESSAYS ON INFORMATION ASYMMETRY IN THE U.S. RESIDENTIAL MORTGAGE MARKET: INCENTIVES AND ESTIMATIONS

Abstract

This dissertation focuses on a phenomenon called appraisal bias in the residential mortgage market that stemmed from information asymmetry. It is composed of two essays, one theoretical and one empirical. The theoretical essay analyzes the existence of appraisal bias in a dynamic game of incomplete information framework and solves for the perfect Bayesian equilibria. It establishes how adding semi-verifiability condition to a cheap-talk game helps construct non-babbling equilibria in an asymmetric information environment. The empirical essay quantifies appraisal bias at individual loan level and measures its effect on mortgage terminations. It tests the extent to which option theory explains default and prepayment behavior in residential mortgage market. It treats default and prepayment hazards as dependent competing risks and jointly estimates mortgage terminations in a competing risk proportional hazard model framework and controls for unobserved heterogeneity using Heckman-Singer nonparametric method. It replaces the inaccurate approximated likelihood function that has been applied so far on competing risks analysis with an exact likelihood function. Armed with repeat sale transaction data, this paper is the first to analyze the effect of appraisal bias on mortgage terminations. It concludes that appraisal bias is important in determining mortgage terminations and needs to be controlled for to correctly estimate termination hazards.