I would like to add on this one that in this particular case, we're looking at differential return profiles, essentially, or outcome profiles of people who were released from the various statuses. It does not control for offence history, and it could very well be that the results you see here are really a reflection of those decisions at the court level, that they got probation because they were of lower risk.
We have looked at data on various risk profiles that are completed when people enter the corrections system, and we do know that the probation population, the conditional sentence population, has lower risk profiles than the custody population. The interaction between the two can't be clearly identified here, but those are other factors that could be operating in terms of the results that you see here.