The Fallacy of the Irrelevant Denominator raises its ugly head again as the soft-on-crime crowd once more tries to play its worn race card.
This time it comes from Illinois, which has a Disproportionate Justice Impact Study Commission. Nobody seems to be interested in making the full report available on the Web, but I did find this "fact sheet" on the site of State Senator Mattie Hunter.
From long experience with people who mislead with statistics, any time I hear "disproportionate" I immediately ask "proportion to what?" A favorite trick among mathematical prestidigitators is to keep your eye focused on the numerator while the skullduggery happens in the denominator.
Sure enough, the fact sheet says (emphasis added): "Nonwhites were arrested at a higher rate than whites relative to their representation in the general population throughout Illinois."
What does this "disparity" tell us of importance to criminal justice policy? Absolutely nothing.
The relevant denominator is people who commit the offense in question. The general population, consisting mostly of people who have not committed the offense, is irrelevant. The ratio of arrests over an irrelevant denominator is irrelevant. A disparity of arrests of nonwhites relative to their representation among offenders would indicate at least a possibility of discriminatory enforcement, but that is not what the report says, at least as far as we can tell from the fact sheet.
On sentencing, they do compare to a relevant denominator of a class of offenders. That raises a different problem. You have to control for other relevant factors, such as prior criminal record. Did they properly control? Can't tell from the fact sheet.
Update: Thanks to commenter mjs, we now have a link to the full report.
This time it comes from Illinois, which has a Disproportionate Justice Impact Study Commission. Nobody seems to be interested in making the full report available on the Web, but I did find this "fact sheet" on the site of State Senator Mattie Hunter.
From long experience with people who mislead with statistics, any time I hear "disproportionate" I immediately ask "proportion to what?" A favorite trick among mathematical prestidigitators is to keep your eye focused on the numerator while the skullduggery happens in the denominator.
Sure enough, the fact sheet says (emphasis added): "Nonwhites were arrested at a higher rate than whites relative to their representation in the general population throughout Illinois."
What does this "disparity" tell us of importance to criminal justice policy? Absolutely nothing.
The relevant denominator is people who commit the offense in question. The general population, consisting mostly of people who have not committed the offense, is irrelevant. The ratio of arrests over an irrelevant denominator is irrelevant. A disparity of arrests of nonwhites relative to their representation among offenders would indicate at least a possibility of discriminatory enforcement, but that is not what the report says, at least as far as we can tell from the fact sheet.
On sentencing, they do compare to a relevant denominator of a class of offenders. That raises a different problem. You have to control for other relevant factors, such as prior criminal record. Did they properly control? Can't tell from the fact sheet.
Update: Thanks to commenter mjs, we now have a link to the full report.
In section 2.3 of the full report, a rare bit of candor in an otherwise result-driven report:
.....suggesting that racial disparities are attributable,in part, to the lengthier criminal histories of non-white offenders.
I see this all the time in studies which begun thusly "minorities are overrepresented in the criminal justice population." This adds the additional flavor of epidemiology as if crime was just a disease like cancer which happens to people.