The statistical confusion about crime rates is off the charts

After the following chart trended on Twitter, it set off a cascade of reply threads from people who had statistically incorrect takes about the issue but were nonetheless very confident.

Actually, let’s start a few days earlier with Mehdir Hasan.

Maher on his show discussed issues regarding the large amount of Black-on-Black crime. Mehdir commented on this, and said the following (1:30):

[sarcastically] The whole Black people killing other Black people. “Black on Black crime”.[/sarcastically]
Are you kidding me? White people kill other white people at almost the same rate that Black people kill other Black people. And yet you never hear anyone complaining on TV about white on white crime.

The paper he refers to shows that for murders commited by whites, ~80% of them are against whites, and ~20% are against Blacks. And for murders committed by Blacks, it’s also true that ~80% of them are against Blacks, and ~20% are against Whites. So these two numbers are similar, yes. But the overall murder rate is much higher among Blacks. Blacks kill other Blacks in extremely high numbers. What he is saying simply does not make any sense.

The weird thing is also that if he thought about what he was saying, he would probably realize that he had confused something. Because he doesn’t really believe that the murder rates among Blacks and whites is the same, does he? But why didn’t he stop himself, and think about this, though? Even after he was corrected by the Community Note, he just doubled down.

But it’s not just him. Then came the interracial crime figure above; and people were equally confused about crime between races, as Mehdir Hasan was about crime within races:

What they are saying here is that there are more possible white victims in the population, and when we account for that, the rates of interracial violent crime are not that different.

It is a correct argument that we should correct for the larger number of possible white victims. But we should also correct for the smaller number of possible Black criminals!

Eg if we look at the proportions of violent interracial crime listed in the table, the number for Black-on-White is merely 50% higher than White-on-Black. But they do this while being a 4.8 times smaller group. So that gives us much higher effective rates for Black-on-White interracial crime. Not “similar” as they both claim.

It is important to realize that most violent crime is committed within race. If we look at the table, 70.3% of violent crime with Black victims are caused by Black criminals. Only 10.6% violent crime with Black victims is caused by white criminals, even though whites are 62% of the population! Because most crime is within race, and Blacks commit that much more crime.

Again here, a weird thing about it is that it’s quite obvious from observing the world that Black-on-White crime is more widespread than White-on-Black. It’s easy for everyone to be confused by statistics sometimes. But once they arrived at this conclusion that the rates are similar, shouldn’t they stop to think “hmm, wait, can that be right? Maybe my logic is wrong somewhere.” But they don’t do that. Instead what they do, is that they make a Twitter thread about it, and get thousands of likes for calling the graph “complete BS”.

The graph is not BS. But if you make a basic statistical error and get some wrong numbers, then of course the correct numbers will look wrong to you.

Kareem Carr also made a massively popular critique, in which he makes essentially the same mistake:

He doesn’t say that the rates of interracial violent crime are similar. Instead he says that it is important to communicate clearly that that the rate of Black-on-White violent crime is about the same as the percentage of Black people in the population. But that is caused by two factors:

1. The vast majority of crime is within race

2. Blacks commit vastly more crime than whites; both within race and between race.

These two very large factors happen to be of roughly the same magnitude, and thus effectively cancel each other out. But the happenstance that they happen to cancel each other out is not a “vastly superior” thing to focus on. In fact, it’s simply confusing, because it obscures the two different factors at work.

Again, you could simply compare it to the 10.6% number for White-on-Black crime, which is much smaller than the 62% White population, not equal to it.

---

Another critique they made in their threads about the interracial violent crime image, is that it is dishonest because it doesn’t show the bars for crime within race.

But that is not dishonest. The graph was making a point about rates of interracial violent crime, and the news coverage of it. It was clearly labelled “interracial”. It is not dishonest to not also show the intra-racial crime, in a graph of interracial crime.

Of course if you think it’s important to show for perspective, you should feel free to show it. But claiming that something is dishonest and misinformation for not doing this, is not a fair critique.

By the way, another thing you could consider for perspective (but they don’t), is that most instances of murder in absolute numbers are actually caused by Black people. 13.6% of the population causing ~60% of homicides.

Indeed, it is plausible that part of the explanation that the rates of violent crime between races are so much lower than within race, is that non-Blacks have fled from the areas where there is a lot of Black crime.

---

An aggravating element to this discourse, is that while the twitter threads containing statistically confused critiques have gotten tens on thousands of likes, the threads correcting them do not get the same level of exposure.

In other words:

And there is no mechanism for correction. The people who made the threads will not read the corrections, and go back and admit that they were wrong.

It’s somewhat ironic that they loudly complain about gross misinformation, misleading statistics, disinformation etc, and have so little focus on trying to get things right themselves.