The Fault Is Not In Our Stars, But In Our Data

Over the past several weeks there has been a great deal of discussion, tweets, emails and concern about what the HMDA data from 2012 is telling us. There are a number of advocacy groups, the MBA and others who are proclaiming that minorities are being denied at a much higher rate than whites. The numbers don’t lie they say and these statistics are leading us to this conclusion. Are you sure about that? Mark Twain has been credited with describing statistics this way: “There are lies, damn lies and statistics.” After all, statistics are just compilations of data and if the data is inaccurate or incomplete, how good can those statistics be? So before we place a lot of faith in these findings we really need to get a much better understanding of just how good this data is.

Recently a compilation of 2011, 2012 and 2013 HMDA data was reviewed. The result of this review found one very profound conclusion. HMDA data is bad. Sure there are validity and quality checks that lenders and regulators use to test the data before and at the time of submission, but the issues go far beyond that.

One of the primary issues is the ability for borrowers to opt out of providing race, ethnicity and gender. Conventional wisdom says that many minorities take advantage of this provision as they are concerned that by indicating they are a minority and fail to report the information. What this actually means is that the number and percentage of minorities approved is likely under reported. So what does this do to the results?

If we want to develop a more in depth analysis of why any group of applicants were denied it is important to have the reasons for the denial. However, some lenders are required to report denial codes and some are not, but it is an option for them. Most lenders not required to do not provide them. Therefore an effective analysis is not possible unless you want to limit it to those lenders who are required to report. Using this limited data set produces a diminished result.

And speaking of denials, lenders appear to continue to have difficulty understanding the difference in the processes between cancelled loan applications and loans denied for incompleteness. Loans denied for incompleteness result after the lender provides the borrower with a notice that the application is incomplete and letting the applicant(s) know that they have a specific period of time to send in the information as described in the notice, or the application will be denied for incompleteness. A cancelled loan application is one where the borrower and/or broker informs the lender, either verbally or in writing, that the borrower wishes to cancel. So are all those denied loans really cancellations or are those cancellations really denied loans?

All of these issues point to one thing and only one thing. The data is bad or at best inconsistent and incomplete. So let’s get a grip on the data before we continue to gripe.

About The Author