Under-specified Models and Detection of Discrimination in Mortgage Lending
by Jason Dietrich
Most empirical studies of discrimination in mortgage lending can be criticized for omitted variable bias. With access to data and policy guidelines typically unavailable to researchers, the OCC is in a unique position to assess the importance of omitted variables on fair lending models. This study examines how variables available to the OCC, but often unavailable to researchers, affect estimates from statistical models and identification of outliers for manual review.
The results show that omitted variables have an important impact on both the estimate of the effect of race and on the identification of outliers for review. Further, there appears to be no consistent patterns to the direction of these impacts. This suggests that it is inappropriate to make generalizations about the potential direction of bias based on assumptions about the correlations between omitted variables and race.
Any whole or partial reproduction of material in this paper should include the following citation: Jason Dietrich, "Under-specified Models and Detection of Discrimination in Mortgage Lending," Office of the Comptroller of the Currency, E&PA Working Paper 2003-2, March 2003.
Any whole or partial reproduction of material in this paper should include the following citation: Dennis Glennon and Amos Golan, "A Markov Model of Bank Failure Estimated Using an Information-Theoretic Approach," Office of the Comptroller of the Currency, E&PA Working Paper 2003-1, March 2003.