[aussenseiter]

Originally Posted by Lune [You must be logged in to view images. Log in or Register.]

The same way they study pretty much anything involving human health or behavior; by looking at recidivism rates between jurisdictions and controlling for variables to isolate cause and effect to a meaningful p value and effect size.

We know crime rates in the 80's went bananas before declining again but we don't really know why. Some jurisdictions went 'tough on crime', others didn't, even within the same city (overlapping sheriff and PD jurisdictions for example, or Los Angeles vs. Long Beach, Dallas vs. Fort Worth). They collated all this data and controlled for things like demographics, poverty, national and local overall crime trends, and looked at recidivism rates for different policies. They compared to similar jurisdictions to control for naturally changing crime rates. Jurisdictions with longer sentences either saw no significant change or an increase in crime, overall.

Malleated data of the seventh order. Why do you trust the conclusion?