[_dps]

While the overrepresentation of the repeat offenders is a valid concern, I feel like this makes a big point out of a relatively small change in conclusion.

  Old Analysis: slightly over half were incarcerated again
  New Analysis: about one third are incarcerated again

So regardless of how you measure it, there's an enormous relative risk of recidivism -- the general population is incarcerated at a mere 0.66% level [1] so that's still a 50x relative risk (compared to a 75x under the old calculation).

There is no avoiding the conclusion that ex-convicts are vastly more likely to perform crimes than the general population. From the perspective of risk management and policy design, 50x vs 75x doesn't seem to change much to me.

And I could forgive a lay person for taking the 50x relative risk to mean "ex convicts tend to return to jail" even if it's "only" 1 in 3. The general rate of incarceration is nowhere near 1 in 3.

[1] http://www.sentencingproject.org/template/page.cfm?id=107

[Camillo]

Well, of course. Criminals tend to commit multiple crimes. Would you prefer a world where crimes are just randomly committed by anyone, including you and your family, with uniform distribution? If you would, you know where to start.

[TheCoelacanth]

Kind of a cop out answer. Obviously some recidivism is caused by some people being predisposed to crime, but when some countries have recidivism rates that are half of other countries' rates, can you really claim that it explains all of it?