They argue that a broad gap between sentence expected and sentence served actually detracts from deterrence, and conclude that "truth in sentencing" laws are good for deterrence (I'm not sure I'd go that far, but it was interesting to pay attention to a differential so far neglected by the literature). [read post]