So you are all prepared to conduct a fair lending analysis. You have determined the focal point(s), to include a product target as well as the prohibited basis and control groups, all of which are topics for another day. Now you are ready to assemble your data and begin conducting the review. In doing so, there are questions you will want the answers to before you begin.
First, what is the sample composition? Is it a single product and, if so, what is the timeframe? This is important because you want (a) the sample to be truly representative of the population of data you are studying and (b) do your target and control groups have an equal chance of being in the data?
Second, are the loans in the sample single or joint applicants? And, if there are joint applicants, is there data on both applicants, and how should these be handled? The answer depends on a few things.
Let’s assume that the product under review contains a mix of both joint and single applicants. The question then becomes how the target and control groups should be formulated for the joint applicants. A number of factors are important in making this determination, including how the institution evaluates applications, such as do primary and secondary applicants receive equal weight, how is the DTI calculated for joint versus single applicants and so forth.
As an example, some institutions have co-applicants shown in the data files that are more co-signors or guarantors than co-applicants and may be treated differently. These types of things have to be understood prior to undertaking the review.
Always bear in mind, whether using statistical techniques or more of a subjective- comparative approach, the sample is only of interest because of what it can tell us about the overall population. In some cases, depending on the data and the loan application pool under review, it may be appropriate to limit the sample to only single applicants. If regression analysis is being used, another alternative could be to operationalize variables into the equations to account for differences between joint and single applicants.
The goal of any type of fair lending review where consistency of treatment is being evaluated is to account for all of the variation associated with the loan files with the exception of the target and control group attributes. An important piece of this is the proper categorization of the target and control groups in the sample. The best approach is to understand that the proper way may be situation-dependent and based on factors such as the distribution of the data, loan product, sample composition, and the objective of the review.
Finally, target and control group categorization may be straightforward once the points above have been determined in samples where government monitoring information (GMI) is available, but not so straightforward when relying on proxies. Having to rely on proxies adds yet another layer of complication and is even more heavily situation-dependent. We plan to expand on this topic in future posts.