5 ms·

100M police traffic stops: New evidence of racial bias

8 points by raiyu 8y ago

zunzun 8y ago
I have read that there are differences in average height between races. Was the height difference statistically controlled for in this study? The article does not mention this.
- consumer451 8y ago
  According to this article[0] the height difference is not significant: > Across the United States, the average black man and the average white man are roughly the same height and weight. According to what data are available, such as information taken from Centers for Disease Control and Prevention surveys, the average white man older than 20 weighs 199 pounds. So does the average black man. Height averages for black and white men are within a centimeter of each other, with the average white man being slightly taller at 5-foot-10. [0] https://www.chicagotribune.com/g00/news/nationworld/ct-male-body-study-20170314-story.html https://www.chicagotribune.com/g00/news/nationworld/ct-male-...
techbio 8y ago
Seriously? They didn't control for the year, make and model of the car? https://openpolicing.stanford.edu/data/ https://openpolicing.stanford.edu/data/
- mlevental 8y ago
  why exactly does that matter? I see a lot people bring things up like this and I always wonder if some feature is perfectly correlated with race, like e.g. car make/model/year, or clothing, or music being listened to during time of stop, what difference does it make? best case the cops are discriminating for the covariates (not something we'd be comfortable with either) and worst case they're using the covariates to discriminate for the confounder. suffice it to say there's no outcome to "controlling" that we would be satisfied with.
  - techbio 8y ago
    No, it would either strengthen their argument or weaken it. Hiding/ignoring the obviously relevant data, it looks like it would've weakened the point they are making, ie. that its persuasive effect has been engineered. If the effect is as profound and as they say (police pull over brown-skinned people because of their skin color and not something else), be convincing if you are convinced of it. Show the data, not just the subset that most cleanly states what you set out to state before analysis. Maybe it shows that black people usually drive Mazdas or whatever and that those are targeted by law enforcement. Maybe it doesn't. We don't know. They've controlled for daylight. Not traffic offense. Not Mazdas. Why not?
masonic 8y ago
Of all of the data sets they use, exactly zero takes the race of the officer or officers into account. If they are trying to claim that statistical differences are primarily explained by racial bias, to completely ignore the commonality or difference in race between the driver and the officers conducting the stop seems like a major flaw to me.

8 points by raiyu 8y ago

I have read that there are differences in average height between races. Was the height difference statistically controlled for in this study? The article does not mention this.

consumer451 8y ago
According to this article[0] the height difference is not significant: > Across the United States, the average black man and the average white man are roughly the same height and weight. According to what data are available, such as information taken from Centers for Disease Control and Prevention surveys, the average white man older than 20 weighs 199 pounds. So does the average black man. Height averages for black and white men are within a centimeter of each other, with the average white man being slightly taller at 5-foot-10. [0] https://www.chicagotribune.com/g00/news/nationworld/ct-male-body-study-20170314-story.html https://www.chicagotribune.com/g00/news/nationworld/ct-male-...

techbio 8y ago

Seriously? They didn't control for the year, make and model of the car? https://openpolicing.stanford.edu/data/ https://openpolicing.stanford.edu/data/

mlevental 8y ago
why exactly does that matter? I see a lot people bring things up like this and I always wonder if some feature is perfectly correlated with race, like e.g. car make/model/year, or clothing, or music being listened to during time of stop, what difference does it make? best case the cops are discriminating for the covariates (not something we'd be comfortable with either) and worst case they're using the covariates to discriminate for the confounder. suffice it to say there's no outcome to "controlling" that we would be satisfied with.
- techbio 8y ago
  No, it would either strengthen their argument or weaken it. Hiding/ignoring the obviously relevant data, it looks like it would've weakened the point they are making, ie. that its persuasive effect has been engineered. If the effect is as profound and as they say (police pull over brown-skinned people because of their skin color and not something else), be convincing if you are convinced of it. Show the data, not just the subset that most cleanly states what you set out to state before analysis. Maybe it shows that black people usually drive Mazdas or whatever and that those are targeted by law enforcement. Maybe it doesn't. We don't know. They've controlled for daylight. Not traffic offense. Not Mazdas. Why not?

masonic 8y ago

Of all of the data sets they use, exactly zero takes the race of the officer or officers into account. If they are trying to claim that statistical differences are primarily explained by racial bias, to completely ignore the commonality or difference in race between the driver and the officers conducting the stop seems like a major flaw to me.