Comment by goialoq

Because correlation != causation.

Here's a very simple refutation:

Is past arrest history a valid factor in the model?

What if past arrest history is biased due to biased policing (Hint: it is, well proven)

How about zip code?