This is a curious approach to the study. You'd think you'd at least consider using census data paired with arrest data. Sure there are statistical biases there,

This is a curious approach to the study. You'd think you'd at least consider using census data paired with arrest data. Sure there are statistical biases there, but they'd probably be smaller than the sampling error.