For many of my gripes with Kleiman, he wrote a book on this topic a few years ago (When Brute Force Fails). Some of the same conclusions are involved. Example. For minor infractions like parole or probation violations instead of sending people to prison to serve out a lengthy term, they might go to jail immediately for some smaller period of time (a few days to a month), and then are released.
A crucial element that wasn't mentioned in the piece is that one of the major aspects of "successful" punishment is that it usually is not the length or severity that matters for people to get the point that they did something wrong and deter others from doing the same. Rather it is having available swift and sure possibilities for penalties to occur in the first place. In many cases prison sentences are increased because we don't have very sure or swift penalties for the criminal acts involved and it is assumed that the only way to provide deterrence is a stiffer penalty. People can commit these kinds of crimes with relative impunity. As a result we might also want to look at which crimes we could better deter with different police patrol patterns, more community involvement, a better focus on the use of our investigative resources to solve these crimes doing the most damage, etc.
People don't tend to randomly kidnap children in the US versus other many countries because it's extremely hard to pull off without getting caught (or at least this is the belief among many criminals, it's also pretty low on the totem pole of prison hierarchies to do pretty much anything to kids, which adds another wrinkle). It's almost always a family member in this country for that reason that abducts a child. Someone who probably doesn't care if they get caught because they're not rational involving their own children. This attitude of deterrence through the surety of penalties could be applied to any number of criminal acts (murder, robbery, fraud, etc) and should provide much more deterrent value than longer sentences. There will still be murders, but they'd be more and more acts of "passion"; sudden events that escalate to violence and less and less random acts of violence committed against strangers.
I think there's also a large sum of currently designated criminal behavior that isn't described or studied here that doesn't cost as much in the form of social damage (drug use/possession and distribution being one source), or where there is damage, using the prison system to deal with it probably isn't appropriate and is wasteful and costly versus the alternatives. This would also apply to many mental health problems. It sounds like this was a study or set of them that looked too linearly at the costs of prison itself rather than also including the costs of policing or the costs of basic health care, and the benefits of same, and so on. I find that many of our policy approaches are not "3-D", taking into account alternatives from outside the context of a narrow policy view. For instance, drug addiction (actual drug or alcohol addiction, not mere use of drugs that we find are socially disapproved) or mental health disorders could be treated as a medical problem rather than a criminal one. The narrow focus on the costs of these in criminal activity often precludes people from assessing how they might be dealt with more efficiently elsewhere. It is useful to know that X criminal act potentially costs us this much in social damage, and imprisoning a person for doing so costs Y dollars, but it is useful to know that because we can then evaluate it against other alternatives and also see if we will come away with fewer of X criminal acts in the first place using those alternatives, or if those alternatives have other positive benefits (better mental health for instance allows people to hold down regular employment easier, narrow policing focus allows police to push criminal activity into smaller and smaller spaces, thus making any remaining criminal activity less likely).