Sunday, August 7, 2016

Raw population and disparate impact example based on college degrees

These numbers explain a lot more than they might appear to do at first glance.



On a back of envelope basis the raw demographic numbers are roughly:
Asians - 5%
Whites - 67%
Blacks - 13%
Hispanics - 15%
Make the rough assumption that most or all professional/managerial roles require a college degree. If that is acceptable as a rough working hypothesis then Hackett's numbers have a buried implication. If you multiply the college degree percentages by the population percentages, you get a professional/managerial workforce that looks something like:
Asians - 9%
Whites - 76%
Blacks - 8%
Hispanics - 9%
Against raw demographics, Asians are especially over-represented (by nearly a factor of two) as are Whites. Blacks and Hispanics are underrepresented by nearly half (17% of the managerial workforce instead of 28% of the population).

On the face of it, a company with this disparate impact between their workforce demographics and the population demographics would invite the question as to whether or not there might be some invidious discrimination occurring, explicit or unconscious. But that would be a false positive. No discrimination at all, simply a mathematical consequence of individual choices.

Disparate impact with raw population as the base has always been a crude and less than useful tripwire to find actual instances of discrimination. As this example shows, disparate impact occurs based on other factors as well.

We ought to pursue actual corporate racial discrimination with some vigor. This exercise suggests that we need a more attuned tripwire for whether possible discrimination is occurring. Too many false positives is a waste of everyone's time and discredits the process when there are actual meritorious cases.

Probably something like Raw Demographics adjusted for Education Attainment, Degree Field, Total Years of Work Experience, Continuous Years of Work Experience. I am guessing that those four variables likely would be very simple adjustments to make and might dramatically reduce the instances of False Positive discrimination claims, leaving the resources available to adequately prosecute real cases.

No comments:

Post a Comment