An examination of 308 studies from the criminal-justice literature, by David Weisburd of the University of Maryland and his colleagues, found that randomised trials were significantly less likely to report positive outcomes than non-randomised studies.
ECONOMIST: Social science