The BISG race / ethnicity proxy model has been used by financial regulators - such as the CFPB - and consumer lenders for almost a decade to detect illegal discrimination in lending outcomes.  But what do we really know about the reliability of these proxy-based disparity estimates?

​

In this in-depth study, I explore this crucial question and provide data-driven results - and underlying intuition - to explain a number of hidden biases that may be driving incorrect conclusions in this high-stakes area.