zlacker

Police brutality is a real issue. It's less clear that racial bias in policing is a statistically significant issue, however, when accounting for obvious things. If someone has numbers that tell a different story I am all ears. Here are mine when I attempted to find it for myself:

    black arrests (all crimes) a year: 2.2 million
    white arrests (all crimes) a year: 5.6 million

https://ucr.fbi.gov/crime-in-the-u.s/2017/crime-in-the-u.s.-...

    black deaths by police: 4.5 per 100k
    white deaths by police: 1.5 per 100k

https://www.pnas.org/content/116/34/16793

    USA black population: 13%
    USA white population: 75%

Given a black committing an average black crime, and a white committing an avg white crime, the black person is 16% more likely to die in a police altercation. Whether or not this is statistical error or a real difference is harder to tell, but this difference is not nearly as large as most media outlets lead people to believe.

Again, If someone has numbers that tell a different story I am all ears

>>randyr+(OP)
The NYC stop-and-frisk numbers should be considered.

https://www.nyclu.org/en/stop-and-frisk-data

>>randyr+(OP)
Laws are disproportionately overenforced on black communities. Your numbers are collected after a large amount of the bias is already baked in and collected by the people perpetrating the bias. Maybe that didn't occur to you.

You're just asking questions right?

>>danhar+41
How do the numbers shake out if we consider only homicides, where selective enforcement should be less of a factor?

>>danhar+41
I welcome more data that has less sampling bias, but without some attempt to quantify the amount of sampling bias here we don’t know if it has a big impact on this data.

>>randyr+(OP)
Wait, what’s your math here? Black deaths by police are 4.5/100k and white is 1.5, that’s a 3x or 300% difference.

>>leeree+B2
This is the report for homicides (a different table from the one in the root comment that tracks arrests): https://ucr.fbi.gov/crime-in-the-u.s/2017/crime-in-the-u.s.-...

Furthermore, this data is from reports not arrests. So selective enforcement is not a factor.

>>manfre+P3
The post above cites "arrests (all crimes) a year", not reports.

>>randyr+(OP)
For all the budding data scientists out there, can you describe your general methodology in more detail?

Here’s my summary:

1. Pick a complex issue.

“racial bias” in policing

2. Pick a (single) year. / Ignore history.

2017

3. Pick a metric, any metric.

“deaths by police”

4. Write conclusion to match your preconceived notions.

5. Congrats, you’re now “data-driven”.

>>awinde+B3
I think it's "30%" (it's not). If p (pct population) is scaled to units of "people", and m as crimes/year and k as (deaths in custody) / (100k people), we want to compare the magnitudes of p * (k / m) for the two groups. Plugging the numbers in as cited, and putting the African American cohort over the White cohort, we get 1.3, i.e. "30%" increase in probability of dying during a police interaction if black.

But as has already been pointed out, it's way more than 30%. The numbers given in the original comment hide the real impact of the bias, since "arrest" was implicitly being treated as a fair event (which it isn't; as just one example, blacks in particular are many times more likely to be subject to a traffic stop than whites, while they tend to have contraband on their possession less often [1][2][3][4]).

Moreover, this isn't just about deaths in police custody. This is about inhumane and repressive policing practices that perpetuate a longstanding effort to deprive blacks of meaningful political power [5]. It is both foolish and cruel to see an entire population struggling and assume it's because they are bad people.

[1] https://sfdistrictattorney.org/sites/default/files/Document/... [2] https://chicagopatf.org/wp-content/uploads/2016/04/PATF_Fina... [3] https://www.aclu-il.org/en/press-releases/traffic-stop-data-... [4] https://www.nytimes.com/2015/10/25/us/racial-disparity-traff... [5] See Michelle Alexander's book, The New Jim Crow

>>randyr+C2
It's presupposing a negligible effect of racial bias on arrest data. That's a very good reason to not use it in showing a negligible effect of racial bias on another scenario involving police. I'm struggling to think of a better example of a circular argument.

>>randyr+(OP)
I'll warn that there really isn't any good data when it comes to use of force. The police departments haven't been required to report it to any national body. That link you have for deaths by police mentions using both official and unofficial sources, and having to fill in missing data.

The FBI has been making some efforts on the issue, but it's quire recent: https://www.fbi.gov/news/pressrel/press-releases/fbi-announc...

>>randyr+(OP)
Looking at murder, the statistics tell a different story — blacks are 13% of the population but over 50% of the murder arrests, while whites are 75% of the population but under 50% of murder arrests.

If times the police use force are correlated to violent crime, then it’s unclear that blacks are over represented in deaths by police — they may actually be safer than whites, once controlling for the distribution of crimes they’re arrested for.

One problem I’ve had in this analysis is that the ~60/1000 deaths per year that aren’t justified (fortunately) aren’t enough to do an analysis on that subset.

Of course, 60 deaths is tragic — but 60 wrongful deaths among 8+ million arrests may not be the problem the media portrays it as.

You’re an order of magnitude less like to wrongfully die from police while being arrested than you are to die from a car crash this year. Overall, police are safer than many things in society.

60/8M // 30k/320M = 8%

>>randyr+C2
It is on you if you don't take the testimony of hundreds of thousands of black people that they're terrorized by the police as evidence of possible police brutality.

Does that sound absurd? Is it as absurd as presuming the police would incriminate themselves with the data they create?

All evidence you can gather will be indirect. You make your judgment based on which you consider relevant.