Fall 2014 question 2c
This might be a hard question to answer as I'm basically asking how graders would treat an answer like this but I'm hoping that I can get at least a decent guess at it. WRT the text message rating variable being discriminatory within a risk classification system, I saw in the examiners report that most people put "equitable" though they accepted "not equitable" so long as proper justification was provided and mentioned that saying it was a bad rating variable was not a proper justification.
My initial instinct with this question before seeing the answer was to say that it was not equitable because it may be that younger drivers text more and younger drivers get into more crashes, so texting would double count the true predictive variable (being young) and thus be unfairly discriminatory.
This seems like I'm riding the line on what they were describing as I'm basically saying it's a bad rating variable, though I think my answer fits properly into the "justifying through a discriminatory lens" framework, but I'm not totally sure. Is there a clearer way to think about this or is my answer something that one would expect to get credit on?
Comments
I don't understand what is meant by equity from the perspective of the individual consumer. I don't find this being discussed in the text. The equity of a texting variable hinges on whether data can be collected for texting while driving.
You are right that this would lead to using two correlated variables in the model, today. It is conceivable that in the passage of time, abundant texting will be a feature of both young and old people.