Not a Lexis+ subscriber? Try it out for free.
LexisNexis® CLE On-Demand features premium content from partners like American Law Institute Continuing Legal Education and Pozner & Dodd. Choose from a broad listing of topics suited for law firms, corporate legal departments, and government entities. Individual courses and subscriptions available.
This is the third post in our "Understanding Fair Lending" series based on a recent fair lending webinar with Jerry Miller. The first article (Fair Lending and Regulation B) is available and 7 more are coming. You can download the full webinar here.
Thanks to the amazing turnout (over 250 banking compliance professionals in attendance) to our free Fair Lending regulation webinar earlier this month, the 30-min Q&A session was peppered with challenging questions. This one in particular asks about the best ways to use HMDA data proxies, the first set of which the FED was kind enough to provide in a webinar last year, and statistical thresholds to watch.
The HMDA proxies provided by the FED are for Hispanic names and were collected from the US census. Thus far, there aren't any other proxies for different ethnicity groups which make the use of proxies somewhat limited. In his answer, Jerry Miller, our fair lending veteran, recommends 200 basis points or higher for a proxy group. Mr Miller also goes on to suggest caution when using proxies specifically for indirect auto lending - examiners will ask "Why is there a 50 basis point difference for Hispanic borrowers in this metro area versus a white borrower?".
In the video below you can hear Mr. Miller answering this specific question, and you might want to read our previous article on the implications of proprietary scoring models for regulation B to have an idea how examiners address these issues.
Of course, this is just a small piece of the 60-min session that took place on 6/4/14. Here are a few more options to supplement your fair lending training:
What do you think? Are there additional fair-lending questions you’d like addressed? Include them in the comments section below or email us at firstname.lastname@example.org. You can also find more Sheshunoff™ training and materials at the LexisNexis® Store.
MODERATOR: What recommendations would you have for the use of proxies for known HMDA data? What statistical thresholds would you use for identifying outliers?
Mr. MILLER: Well, first of all, good point. This came out in a presentation by the Federal Reserve System in 2013. In fact if you look in their consumer affairs porthole and look at the webinar that was provided, I believe, through the Philadelphia Federal Reserve Bank, you will see that the FED has very kindly provided you a list of proxies for Hispanic names.
Unfortunately it's a little more difficult to use for other origins, ethnicities because we don't have that by name. However, majority-minority census tracks is another methodology for proxies.
What I look for in terms of difference when I use proxies is if . . . let's say denial rate. If the denial rate, and historically I use 200 basis points with higher for a proxy group, I would raise my eyebrows and ask the question "Why?" Today, one of the areas that we have seen proxies use is indirect auto lending and when they see differences higher than even 50 basis points, regulators are going to ask the institution “Why is there a 50 basis point difference for Hispanic borrowers in this metropolitan area versus a white couple, a white individual,” whatever the scenario might be used for the control comparison. I hope that's helpful. That's based upon experience without getting into actual drill-downs for a specific case.