To use the paper's results, it looks like they're getting roughly 10 in 100 in the experiment condition and 18 in 100 for the control. Those kids were selected because they were considered high risk. If among the 82 of 100 kids who didn't get arrested there are >18 who are just as likely to be arrested as the 18 who were, then emile's conclusion is correct across the year. The majority won't be arrested next year. Across an entire lifetime however.... They'd probably become more normal as time passed, but how quickly would this occur? I'd think Lumifer is right that they probably would end up back in jail. I wouldn't describe this as a very regular problem though.