Feature wish list

[norcalbiostat]

If there is such a wish list growing I would love to see the following characteristics added to the data.

• Officer race
• Officer age
• Years of experience
• Did the deceased have a prior criminal record?

[minorsecond]

Perhaps these as well (I'm currently working on it):

• Population by race, in each city with a shooting
• FIPS code for Census place (make relating your data to Census data easier)

[kibiz0r]

I'd like to see the status of legal proceedings as well. Mapping Police Violence is able to get that data somehow.

[lamahmud]

I believe years of experience is the most telling field we need added. I have a theory that most officers involved in shootings suffer from PTSD. PTSD creates hyperawareness and heightened reactions to surprise stimuli. In my cursory googling of officer involved shootings, i found that most officers aren't rookies. They are usually 3 years or more in. That is way too long to be a patrol officer. Even the military rotates troops off combat/patrol duty. Being a patrol officer is 99% boredom and 1% terror. Unfortunately, that 1% of terror can happen during any interaction so cops end up hyper alert after a while. ie PTSD.

[kanhema]

That's an interesting observation @lamahmud –– Definitely worth looking into.

[ghost]

I'm also very interested in more stats/perspective into the officers involved. So I automated google searches for each person in the list (using Google API), got the top 3 results for each and then automatically downloaded youtube vids, police report pdfs and article htmls, using wget and youtube-dl. This gave me about 8GB worth of content that I can now [more] easily sift through. More plans to automate finding keywords like 'aquitted', 'officer XXXXX' etc.... Then maybe automate searches for their names to find out years of service and other details. I can link to part of a download to that content is anyone is willing to help.

Project is here: https://github.com/BiTinerary/anEmptyKitchenChair

Obviously if I'm able to gather more info and pertinent columns I'll post publically. Try and do a pull request or w/e here.

[obxkid413]

Agree that looking to see any legal proceedings information for the involved officers would be interesting. Years experience would be good to know and maybe police department name or precinct would be good as well.

[adamdill]

Another +1 for FIPS code, address, or any more granularity on the location of the shooting. If we had that data, we could study socioeconomic effects on shootings. Only providing a city isn't granular enough to come to real results on that front.

[servusdei2018]

@washingtonpost I'd like to see these stats:

• Whether the shooter had been using drugs. If so, which?

[rudi3838]

I'd like to know the crime the police was called for (or the reason the patrol stopped) .

[skendrot]

More information about the officers involved would be great!

[JohnJChristie]

Please also add victim SES. I'm quite certain that the whole set is poor people but it would be great to see actual SES.

[andret6]

Another +1 for incident location, would be happy with offset coordinates if that is more feasible for privacy concerns that street address

[jmuyskens]

We have now added the location coordinates to the data.

[lordneeko]

Perhaps these as well (I'm currently working on it):

• Population by race, in each city with a shooting
• FIPS code for Census place (make relating your data to Census data easier)

This data is missing the total number of officer interactions by race and location as well. Without knowing the total number of interactions, you lose context of the data when comparing between different categories.
For example, if there are 4 police shooting of blacks and 2 police shooting of Hispanics. Then you cannot conclude that police shoot blacks twice as often as Hispanics unless the number of interactions between the two is equal.
However, if the number of interactions is 2 to 1, then the actual RATE comparison between blacks and Hispanics, would be equal.
On the other hand, if the ratio of the number of interactions is 1 to 2 then the Hispanics rate is actually WORSE than the blacks (All example data obviously) even though the sheer number looks worse. This would make any reporting on the numbers alone (or by total population) as false or misleading information.

[daledali]

Location (ie in a hospital) and if the victim was a patient.

[CuirPork]

The current narrative that your data is being used for is to compare these numbers to the national population distribution. If we were being honest about this (instead of just selling subscriptions), we would have a breakdown of the police interactions by race and how many didn't result in death. Simple, low-ball estimates of 5 days a week, 5 interactions per day, 50 weeks a year for 600,000 cops should seem to indicate that out of 750,000,000 cop-related actions per year to the US (that is if cops only had 5 interactions a day each, only worked 5 days a week and took 2-weeks vacations) the percentage of those that end in death (roughly 1000 for years before the protests started) means that 1000/750,000,000 people were being killed by police for the years leading up to the protest. That's about 0.000001333333 or .000133% of all cop interactions in a year end in death.

I realize it is futile to report this information because it would be hard to compile, but mostly because it completely invalidates the hype you have been so carefully cultivating these years and because you would lose subscriptions when people found out the truth., but for the sake of scientific accuracy and rigor, it would be nice to have this information.

Oh, but wait. We can figure that out using the other common claim made by activists. Cops are 9x as likely to interact with a black person (stop or patrolling neighborhoods) than a white person. Swap those numbers in rather than the racial distribution of the entire US (which is ridiculous to consider) and let's compare the racial distribution of cop interactions (90-10) to the racial distribution of people killed in the line of duty (24/64). THAT WOULD GIVE YOU ACCURATE estimates, now wouldn't it?

[mikekeith52]

Sir, this is a GitHub.