Studies and Data

The other day AVI posted a link, which I found very interesting, to a refutation of a prominent study allegedly showing that police were more disrespectful to minorities. In fact (and this perfectly mirrors my anecdotal experience), the police are almost perfectly formal with everyone in their professional encounters.

Of course, 'disrespect' isn't really what's at the back of the current dispute; what's at issue is people getting killed by police, not people being subject to rude language by police. Vanity Fair has compiled 18 sources of data that all seem to point in the same direction on that.

Some of these sources are better than others. All of them may be subject to the usual problems of confirmation bias, and the fact that people in the academy really want to prove racism (and may, indeed, fear for their careers if they seem to disprove it). I understand all that; but some of these findings are worth noticing.

This one jumps out at me above all:
3. An analysis of the use of lethal force by police in 2015 found no correlation between the level of violent crime in an area and that area’s police killing rates.
Numbers one and two establish that unarmed people are much more likely to be killed if they are black; that's of small concern to me, since I'm typically always armed, but it suggests that non-black Americans have more leeway to 'opt out' of violent encounters with the police.

Number three, though, that's astonishing. It's completely counter-intuitive. But here's the chart:

It seems like there ought to be at least some answers in all this data, at least that part of it that looks reliable on examination. I'm inclined to continue to favor the hypothesis that training is largely at fault, as I have argued in the past, because it could in theory account for this strange lack of correlation between violent crime rates and police killings. If they're being trained to resort to guns in the face of certain stimuli, then a number of considerations related to an in-context analysis of how dangerous an environment really is may drop out of the 'shoot/no-shoot' decision.

In any case, a look at the data is more hopeful than another round of 'hey, let's hate each other' shouting. Take a look. Maybe you'll see something that helps.


douglas said...

Looking at the table, one possibility that comes to me is that many of the large gap cities (gap between crime rate and deaths) is that they likely have a large number of non-English speaking stops. Purely a WAG though.

One problem I think is that when you're talking about police shootings, it's a pretty small number in the scheme of a country of 330 million and I think somewhere around 85,000 police encounters annually. One of the studies in that linked piece was ridiculous- cop on off-duty cop shootings which numbered 11- and they drew a conclusion from that? Not scientific. Of course, never trust a piece that just gives you clips cut out of context and nothing else, right?

douglas said...

This also caught my eye (from the linked piece):
"The researchers also found significant differences in the way officers spoke to African Americans: “Using only the words an officer uses during a traffic stop, we can predict whether that [officer] is talking to a black person or a white person” with 66 percent accuracy."

Now, if that study didn't in some way control for how the officers were spoken to, those numbers are meaningless. I'd suggest they reflect the cultural differences between how blacks perceive cops vs. how whites do, much more so than telling you anything about the cops beyond that they're human.

Christopher B said...

Sort of like the quip that the common factor in all of your bad relationships is you, once you get past the 'racist cops hunting minorities' BS every one of the incidents has a cop who felt that deadly force was needed to protect their life or the lives of others. The useful question to answer would be why did they feel that way?

Grim said...

That's a good question.

One of the stats, from the same study, is that 99% of these officers were not convicted of any crimes. So, one possible answer is that occasions where deadly force are warranted don't correlate with violent crime rates; they come up sort of randomly, and the police are just doing the right thing (at least 99% of the time).

That would be kind of surprising, but it's not impossible. We'd expect to see that if a further study showed that these incidents tended not to be connected to ongoing criminal enterprise. That might even be possible: fewer than one in three of the black people killed by the police were suspected of any crime, or of being armed, at the time they were killed. (Same link as the chart).

On the other hand, that seems to be counterindicated by the fact that Oklahoma has a rate of these killings 7x greater than Georgia; if it were random, you'd expect it to occur in more populous states more often. It could randomly occur 7x more often in Oklahoma, but that would be surprising; random things should occur more or less as often to one person as to another, so where there are more people, there should be more things.

Another possibility is that (as I've suggested) it has something to do with the training. Then the lack of correlation to crimes can be explained because the crimes are irrelevant; what's causal is the training. (That also accounts for the 99% acquittal rate: if people are following the training, of course they're acquitted because they did what they were told they were supposed to do.)

And that has some support too; again, same link, police departments that have adopted alternative use of force policies kill "significantly fewer people." But since few departments have adopted them, that's not dispositive.

douglas said...

Good points.

Looking at it again, I also noticed that the killings index is per 1,000,000. That's going to make it difficult to compare Bakersfield (<400,000) to Los Angeles (almost 4 Million). One police killing in Bakersfield is three on that graph, while in Los Angeles, three killings is less than one on that graph. If you took that data and resorted by city population, I bet it would look quite different.

Agree that it's unfair to compare cities that have completely different policies on the use of less-than-lethal weapons. There are many variables at play here.

Also, consider this:
"In 2011, over 62.9 million U.S. residents age 16 or older,
or 26% of the population, had one or more contacts with
police during the prior 12 months"

From USDOJ report

Total deaths by being shot by a police officer, on duty or not, justified or not- is less than 400.

That is such a small number, that would indicate that it's going to be difficult to find a pattern in what are essentially extreme outlier events by nature.

Grim said...

That's true.

At AVI's in a parallel discussion, somebody flipped the data: what about police killed by blacks vs. others? It looks like blacks represent about 4 in 10 cop killers, at 13% of the population. But again, how many police are killed? Not many! 126 out of 1.1 million, or 0.011%.

So when we're dealing with a situation in which 99.989% of the police are omitted from the statistics, it's going to be hard to make decisions about how the statistics shake out.

It's a good thing that these are 'extreme outlier events,' of course. That's one place where we are getting things right. The last thing we should want is more reliable statistics to arise from a much greater incidence from which to calculate.

james said...

Where are the error bars on those plots?

Texan99 said...

No pattern of any kind jumped out at me, but I have to admit I was surprised to see cities like New York and Philadelphia down on the very low tail of the graph. But Oakland and many California cities were on the high end. No idea.

Grim said...

"Where are the error bars...?"

Clearly they aren't there. It's a good point.

This morning I've been wondering if the failure to correlate turns out to be a function of the use of city data, and not neighborhood data. Not that I insist that there must be a correlation between violent crime and police shootings, but only that it would make so much sense that its absence deserves to be proven before accepted.

I wonder if we could recreate that data on a neighborhood level? My sense is that the US murder rate is very low outside of certain major cities, and also that even within those cities, it is very low outside of certain specific neighborhoods. If the police shootings tend to occur in those areas, then we'd have a more intuitive picture. If not, then it really isn't about actual crime at all.

douglas said...

I wish I had the numbers as well as the graph, as then it would be relatively easy to reconstruct the graph, then rearrange it different ways to see what happens.

Another problem with this graph, related to what I pointed out earlier, is that the vertical axes, while both regarding per population use wildly different scales. Police killing in per 1 million, while the violent crime is per 1,000? Imagine what that graph would look like were those scales proportioned accurately to each other?

Look, whatever the case about the individual incidents- whether they are legitimate uses of force or not- one thing that is crystal clear to me even more now is that there is nothing close to an 'epidemic' of cops killing blacks.

In the United States today, your odds of being killed by a cop (illegitimately or legitimately, all races) are on par with the odds of being struck by lighting.

Black Males are around 14 times more likely to die at the hand of another black man than at the hands of the cops (legitimate use of force or illegitimate use of force). This of course would be dramatically higher if we could accurately limit it to legitimate uses of force, but that's at issue here, so I'll use the worse number.

And this is all with numbers for 2009-2015. It'd be more exaggerated now with violent crime increasing.

Gringo said...

On the other hand, that seems to be counterindicated by the fact that Oklahoma has a rate of these killings 7x greater than Georgia; if it were random, you'd expect it to occur in more populous states more often.

My only encounter with Oklahoma police- a highway patrolman- did not leave me with respect for them. I was driving after sundown in the country about a half hour short of my aunt and and uncle's house. I stopped to pee on the side of the road- about 10' from the road. I finished and walked towards my car. A police car pulled up. The policeman asked what I was doing. I replied that I had just peed. The policeman acted as if I had exposed myself at high noon in downtown Oklahoma City. He was really upset. Looking at my driver's license, he informed me that he didn't like my "east coast attitude." I thought to myself that my Okie grandfather could have been farming the field adjacent to that road. If he needed to pee, he would have done so on the spot. That's what you do in the countryside.

He didn't like that I addressed him as "Officer." He was a "Highway Patrolman." Or something like that. He was a young guy who probably hadn't been born when I made my first visit to my Okie grandparents. I wonder how long he lasted on the job.

Grim said...

I had a Georgia Highway Patrolman stop to ask me what I was about once for similar reasons. He didn't give me a hard time -- he just said, "Oh, well, be sure and get far enough off the road."