selectivity bias and CDC report on juvenile transfers

the washington post and other media have publicized a new CDC panel report published in morbidity and mortality weekly. after comparing recidivism rates in six strong studies of youth transferred to the adult system with those of youth who stayed in the juvenile system, the authors conclude the following:

Review of the effects of transfer laws on subsequent violence indicates that the experience of transfer to the adult criminal justice system is associated with subsequent violence among juvenile participants when compared with violence among juveniles retained in the juvenile justice system. In addition, little evidence supports the idea that transfer laws deter juveniles in the general population from violent crime... use of transfer laws and strengthened transfer policies is counterproductive to reducing juvenile violence and enhancing public safety.

hmm. though i'm sympathetic to the authors' viewpoint and i really liked each of the studies cited in the report, i'm not completely convinced that they have cracked the problem of sample selection. this is a very difficult thing to do in this research setting, since kids are (literally) selected for transfer on the basis of their perceived dangerousness and likelihood of recidivism. here is the relevant passage on selectivity:

All of the included studies attempted to control for possible selection bias by restricting the cases under consideration to serious ones that would be eligible for transfer and by comparing the outcomes of cases transferred with those of cases retained in the juvenile system. In addition, they attempted to reduce selection bias by one of three methods: 1) by using statistical methods to control for factors that might affect transfer decisions (23–25); 2) by matching transferred and retained juveniles on background characteristics (26,27); or 3) by comparing the outcomes of juveniles matched on background demographics, economics, and crime characteristics, but in jurisdictions with difference transfer laws (28).

well, that's a good start, i suppose. what were the results? of the six studies of transfer to the adult system, one found a deterrent effect, one found no effect, and four found widely varying estimates of increased violence or general crime. the cdc report did not discuss the suspected mechanisms for the deleterious effects of adult transfer, though i believe that the literature typically offers some variant of a brutalization hypothesis.

my sense is that transfer to the adult system probably does indeed increase recidivism and compromise public safety. that said, the specific selection criterion to get into the treatment group in these studies (predicted dangerousness) is uncomfortably close to the substantive outcome measure used to assess their effectiveness (violent recidivism). that's why i'm not sure that the evidence is strong enough here to warrant definitive causal claims. perhaps it is safer to state the conclusion in the negative: after examining the best available studies on the subject, there is almost no evidence suggesting that adult transfer provisions reduce subsequent crime.