A question that often arises in fair lending regression analysis of lender loan decisioning practices is how special case loans are handled.
These are applications that may have been underwritten differently than other loans in the sample under review and, thus, the outcomes were determined by factors outside of the model.
The question becomes, should these loans be simply excluded from the analysis or should they be retained and accounted for in the modeling somehow? The circumstances of special case loans may vary as do opinions on how they are to be handled. We address this question in our white paper Special Case Loans: Should We Exclude or Control for Them?
In this experiment, we test different methods and assess the impact of each on the accuracy of the regression results.
The paper can be downloaded below.