In Part 1 of this post I raised the issue of the new HMDA to be reported March 1, 2018, and how this has the potential to alter the fair lending landscape by increasing the level of risk.
Below I detail some of the reasons why I believe this is the case. In a future article, we will discuss steps that can be taken to be prepared and lower exposure to the increased risk .
With regard to the new HMDA data to be reported, it is important to first keep in mind that agencies have taken a statistical approach to evaluating fair lending compliance. This approach uses sophisticated econometric techniques such as regression analysis. I remain surprised at how many lenders with significant HMDA volume still have no idea what their data would look like if subjected to such an analysis.
Even more lenders fail to recognize the complexity of such methods and rely entirely on either “canned” software solutions or otherwise abbreviated methods that may not be telling the whole story. I can cite numerous examples in which such analyses failed to reveal significant disparities that were subsequently discovered by regulators.
Cutting corners and choosing the least expensive alternatives can be extremely costly in the fair lending world.
Second, the data reported is precisely the same data that would be used in a fair lending regression analysis. Although one could argue that this data does not include every factor; and, therefore, any analysis without additional data would be inconclusive, the reality is that these data account for the majority of the variation in both underwriting and pricing. Agencies as well as advocacy groups will be able to conduct regression analyses that formerly could be done only onsite or after requesting data from the institution. I don’t believe the risk this poses can be overstated.
Third, by virtue of the public data, relative comparisons can be conducted between institutions. So any analysis is no longer limited to individual institutions, but comparisons on an almost infinite number of measures will be possible.
Fourth, related to the above, outside of regression, numerous simple calculations could be done to assess and compare lender practices. For example, comparisons can be made of denial rates for protected and non-protected groups for borrowers with certain credit scores or DTI ranges. In the past, these data were not available so such comparisons were limited to only the group designations itself without qualification. Even though such analyses may be incomplete, they could still be perceived as valid and could reflect negatively on the institution and invite further scrutiny.
Finally, there may be products an institution will be reporting that have not been part of HMDA in the past. One example is HELOCS. Even if an institution has done sufficient analyses of its HMDA, products that have not been part of HMDA in the past have not been analyzed with GMI. Therefore, many lenders are likely unaware of what these data may reflect when subjected to statistical analyses with GMI.
In summary, I firmly believe the HMDA reporting change greatly expands fair lending risk for financial institutions. Regardless of whether this prediction materializes or not, lenders would be wise to heed the warning and begin taking steps now. This all can, however, be turned into a positive. Watch for our upcoming article, HMDA 2018 – How Statistics Can Be Our Friend, in the coming weeks.