Bad Statistics About Bad Things

Critics of policing and mass incarceration make a basic statistical mistake.  Men represent less than half the population of the country but over 90% of those incarcerated.  But no one says the system is sexist.  That is because men commit more crime.  Comparing population numbers to incarceration rates gives a false picture.  Crime rates should be compared to incarceration rates.

Why are groups imprisoned at higher rates than their representation in the population?  The first question must be whether these groups are doing things that get you imprisoned at higher rates than their representation in the population.  Similarly situated groups can be compared. Groups that are not similarly situated cannot be compared. In America today, ethnic and racial groups are not similarly situated. There are many reasons for this, including the systemic racism of the recent past. The Jim Crow laws did not predestine anyone for a life of crime, but they didn’t help either. And there are many more reasons for ethnic and racial difference that are worth knowing. The rich and the poor are not similarly situated. I’ve read that the most predictive factor for criminality is age, then sex. Social scientists try to control for variables like relative crime rates. We should do the same, even though it is a difficult conversation to have.

Unfortunately, for many, the analysis goes something like this: look at the percentage of the population that belongs to the group, look at that group’s representation in prison, and draw a conclusion. Social scientists know better. The media should know better, and so should we.

Here’s how Crime and Consequences describes the problem:

One thing we do not need is the exaggeration of the prejudice problem created by all the bleating about “disparities” in statistics based on the Fallacy of the Irrelevant Denominator. Comparisons of the demographics of Statistic X with the general population are irrelevant if offending rates are not uniform across the demographic groups in question, which they rarely are. Do the police ticket more men than women for speeding? Of course they do, because a greater percentage of men speed. That statistic does not establish or even indicate discriminatory enforcement unless one establishes that speeding men are more likely to be ticketed than speeding women. The general population is an irrelevant denominator. The people who commit the offense in question make up the relevant denominator. And so it is for criminal justice statistics generally.