In a recent post, "a quick check on long-term violence in the US," we posted the following graph:
You see the lines are fairly similar, though killing of police offers drops below police killings, where police killings per capita remains somewhat constant.
We noted originally that the officer death rates were per 100k, and the police justifiable homicide rates were per 10 million. We concluded from this that police were killed at a rate of 100x the rate that they killed other people.
This was dead wrong.
For two reasons:
First, only about 50% of police deaths are from a homicide of some sort. Others are crashes and other line-of-duty hazards. So that cuts the number in half.
But the other thing worth noting is that there are about 800k police officers, and about 242 million adults in the US. This means there is about 1 police officer per 300 adults in the US.
So the rates actually change dramatically. We divide the 100:1 number by 2 (for half homicides) and then by 300 (for population scale), and the killing rate changes.
Police actually kill people 6x as often as they are killed by other people. So I was off by a rate of 600.
(This pointed out by reader Dan; thank you.)
It actually took me a minute or two to figure this out. I'm not great at statistics.
How can you figure this out quickly? You can use an extreme edge case.
Let's say that there were only 10 police officers. They kill one in a million people, so that's 2.42 people. They have a 10% death rate, so one of them is killed.
It would mean that the death rate of police officers is 1 in 10, and the death of citizens is only 1 in a million. So does this mean that they're a million times more likely to be killed, than to kill? No, not at all: they're 2.42 times more likely to kill than be killed.
When you're fact-checking, those head-checks are really important to not blow it by a factor of 600.
Why is this important?
Whether there is one sort of problem or another, the numbers do matter. Data isn't everything--you do need to think about what you really care about, what your values are--but when you want to help people understand the scale of a problem, and whether it should be your priority, you gotta get your numbers right.