While I can't give you a link to this study, I can say that it supports and extends work I did as part of a graduate stats class for my MBA - we looked at the data that was publicly available at the time and it was clear that the cameras weren't making a statistically significant difference at most intersections... So, just another piece of anecdotal evidence but if you really want it I can try to dig up our work.

Retrospectively, I guess we should have been better about sharing our findings.