Fair Lending on Non HMDA loan data

Posted By: JustGottaKnow

Fair Lending on Non HMDA loan data - 04/10/12 06:55 PM

I'm in the CRA dept and our Compliance dept is handing off fair lending on non hmda items to us. Outside of FL Risk Questionaires, no actual FL analysis or file reviews have been done. I was hoping that I could get guidance on what the FL requirements were for non hmda data and if anyone has already established procedures for this type of FL review.
Posted By: swiggles

Re: Fair Lending on Non HMDA loan data - 04/10/12 07:33 PM

You'll probably have to review based on apparent Hispanic and Non-Hispanic surnames within each product (so that you're always comparing apples to apples). You can do a rate comparison and then analyze variances. Also look at the approval time between application and action taken and look at denials vs approvals. That's my tip-of-the-iceberg two cents. I'm sorry I don't have any established procedures and don't have a huge amount of exposure to this. If loan officers have discretion with respect to pricing and approvals, you're going have a difficult time.
Posted By: Many Hats

Re: Fair Lending on Non HMDA loan data - 04/10/12 11:34 PM

Excerpted from this link:


"...For nonmortgage loans, we may determine ethnicity and gender using the U.S. Census Bureau's Spanish surname list and female first name list. For both mortgage and nonmortgage products, we also use census data to identify majority-minority census tracts and to determine whether disparities exist between minority and nonminority areas."

The entire issue of the 1st QTR 2012 Consumer Compliance Outlook is highly suggested reading!
Posted By: swiggles

Re: Fair Lending on Non HMDA loan data - 04/11/12 03:28 PM

In an exam, keep a careful eye on that method. All Hispanic surnames are not in that list. You (or your loan officers) know which customers are Hispanic. Or perhaps which customers have Hispanic surnames due to marriage but are not Hispanic. I think that method of testing for fair lending ought to be outlawed because it's not provable.