All of the acceptable studies in the literature index the reoffence rate according to several things. One is whether he was convicted of another offence. Was he charged with another offence, but the offence was withdrawn? Did he break conditions of parole? All of those we count as a reoffence. That's over-counting reoffending. Still, in this large-scale study, we have 5.2% recidivism in the treated group and 18% recidivism in the untreated group.
On March 24th, 2011. See this statement in context.