Within bounds of statistical variance, this is an interesting idea. I'm thinking of small sample sizes and the negatives of publicly posting an algorithmicly-derived statement of, effectively, racism. And do we normalize against those factors being transparent to renters? (If I reject an applicant with no photo who happens to be black, does that "count"?)

But if you get it right, is there any legitimate explanation for why statistical differences would exist? Or do you baseline against the entire AirBNB population (which may be biased in of itself)?