The conventional segment of the mortgage market has changed a great deal since the 2007/2008 financial crisis.  This study examines the changes in the Government-Sponsored Enterprise (GSE) mortgage underwriting regime in order to understand the impact of these policy shifts on borrower access to credit.  The analysis is based on the universe of home loan applications processed by the GSEs' automated underwriting systems between 2005 and 2012.  Because mortgages guaranteed by the GSEs receive an implicit public subsidy in the form of lower interest rates, it is important to understand how the GSE underwriting outcomes are related to borrower and loan characteristics.  How did the population of borrowers and loans that received the approve/accept rating by the GSE underwriting systems compare to those that received the adverse risk class designation?  What borrowers and loans types were served by the GSE market?  We look at the distribution of borrower and loan characteristics by underwriting outcome.  A particular focus is the comparison of outcomes before and after the burst of the housing bubble.

Looking at the changes in the composition of loan applicants over time also shed important light on how the GSE underwriting regime has changed.  Evidence suggests that the conventional mortgage market tightened substantially in the years after the financial crisis.  However, the mechanisms of this tightening are poorly understood.  What borrowers and loan types were impacted the most by the underwriting changes?  What explains the changes in GSE loan approval rate over the years?  To answer these questions, we develop Logistic regression models to understand how loan approval rates are related to borrower and loan characteristics.  Model parameters are used to calibrate and compare the expected loan approval probability for a prototypical set of loans from two time periods: before and after the financial crisis.

We find that the changes in underwriting standards did not impact all borrowers equally.  All else being equal, home-purchase mortgages with a high LTV ratio and borrowers with a less than perfect credit score (i.e., FICO below 710) faced a significantly higher chance of receiving an adverse underwriting outcome in the post-crisis era than in the pre-crisis era.  Borrowers with a good credit standing tended to find it easier to obtain credit after the financial crisis than before.  For homeowners trying to refinance, their expected probability of loan approval was actually higher in the post-crisis period than in the pre-crisis period, after controlling for loan and borrower characteristics.

In the final part of the paper, we employ a decomposition technique to examine the changes in the GSE loan approval rate across the years.  It reveals that, aside from the changes in the underwriting regime, a significant portion of the observed difference in loan approval rates between the two time periods was attributable to changes in borrower and loan composition in the pool of applicants after the financial crisis.  These compositional shifts most likely reflected a combination of borrower self-selection and the increase of lender overlays and pre-screening of loan applications in the post-crisis era.