Scottish poet Andrew Lang said “Most people use statistics like a drunk man uses a lamppost; more for support than illumination.” Judging by their behavior, right-wing pundits have taken that as a binding instruction. Wingnuts are constantly using rhetorical games to obfuscate their real intention, and misuse of stats is one of their favorite techniques.
Today’s example is this pile of garbage produced by Richard Hanania, one of the newer and smaller fish in the wingnut universe. Hanania repeats that greatest of right wing hits, the assertion that Black people are more prone to violent crime, and that the solution to crime is more cops and more prisons. He bases this assertion on one single bar graph give in this Tweet. Based on this assertion, Hanania claims that the solution to crime is “You need more cops, more prisons, and more use of DNA databases and facial recognition technology.” So how do we analyze this assertion? How can we determine if any particular statistic is valid or useful, and not simply the Drunk’s Lamppost?
Statistics are really just narratives expressed as numbers. They’re not, in themselves, evidence of anything. They can show comparisons in easy formats, but they don’t prove that the comparisons are valid. In this case, Hanania uses numbers taken from the 2018 Statistical Brief of Race and Ethnicity of Violent Crime Offenders and Arrestees. This is a report compiled by the US Department of Justice based on two sets of data, the FBI’s Uniform Crime Reporting Program (UCRP) and the National Crime Victimization Survey (NCVS). Whether Hanania’s graph is useful depends on whether the information contained in those two reports is useful, and the evidence shows clearly that it is not for the purpose he proposes.
First, we need to define terms, so we’re all talking about the same things. “Crime” as I use it means “actions specifically banned in state penal codes which subject those who do them to incarceration.” This means that “crime” is necessarily a social construction, as it is something that legislatures define. United States v. Hudson and Goodwin, decided in 1812, stated “The legislative authority of the Union must first make an act a crime, affix a punishment to it, and declare the Court that shall have jurisdiction of the offence.” It should surprise no one that legislatures often impose their own bigotries on criminal law. Still, ‘crime’ has a specific meaning as something defined by a state legislature.
So how accurate are the two reports used to create that statistical survey?
The first fact to note is that the reports don’t follow all the material included from reports to convictions, and the NCVS doesn’t even restrict itself to crimes reported to the police. None of these reports have been subjected to any kind of review to see if the people making them were truthful or reliable. Their results can be summarized by saying ‘this is how people responded to a survey about crime’ (NCVS) and ‘this is what a bunch of police reports said’ (UCRP.) It is not possible to determine from the information provided whether the victims knew whether what happened to this was an actual crime or not, because we don’t have access to the original interviews or police reports.
The first objection to my last paragraph should be ‘people know if they’ve been robbed or beaten.’ This is true, but there are other factors that affect the conclusion that something is a crime, and without a complete record, we can’t know those factors. (One obvious example is whether the other person was defending herself, which we can’t know just from hearing one side of the story.)
The fact that the UCRP and NCVS are based on self-reports by victims is much more important in analyzing its claims about race and crime. The NCVS bases its offender demographic characteristics on the victims’ perceptions, a fact noted in the 2018 Statistical Survey. If the person making the report is mistaken — or malicious — their information makes the final statistics less reliable. No one here sent DNA samples from all the alleged perps to 23AndMe for a full analysis, so we’re stuck with the reporters’ impressions based on what they saw.
The problems with perception is ameliorated to an extent by the fact that many of the victims knew the perps. Even then, there are reasons to question the reports. As the Statistical Survey says on Page 3, victims in 67% of nonfatal violent crime incidents “said they knew all or some of the offenders or had seen them before.” The inclusion of ‘people the victim had seen before’ as the equivalent of ‘known to’ the victim means that some of those reports were based on the victims of assumptions of the race of the perpetrator. It is entirely possible that the reports erroneously state the race of the perpetrator and, again, there’s no way to check the accuracy of the report. In incidents with multiple offenders, only 34% of victims said all of the offenders were known, “18% said some were known, and 48% said all were strangers they had never seen before or said they did not know.” This means that 48% of the scariest incidents — multiple offenders attacking one person — were unknown to the victim, making this statistic open to very serious questions. If the alleged offender had black hair, brown eyes, and a dark tan, the victim could have assigned the offender’s race to about 9/10ths of the human race.
Hanania’s use of one report to recommend brutal and expensive policing is especially egregious since he ignores the one, absolute, and inarguable genetic trait associated with committing crimes: the Y chromosome. In Texas in 2021, there were 133,772 people incarcerated in total and 123,263 of those were male. That’s pretty easy math. If we were really serious about preventing crime, we need to focus on men regardless of race. Of course no one on the right wing ever recommends that. They always lean on the Drunk’s Lamppost while pontificating about how the problem is always black people.
