Death penalty as a deterrent?

This NY Times article discussing a fair amount of evidence that the death penalty does actually function as a deterrent for murders is unsettling for a luke-warm death penalty opponent like myself. The summary is that there's a fair body of evidence that it prevents murders, maybe 3 to 18 murders per execution, but experts disagree over whether it's conclusive.

The implication if it is a deterrent is in the article: “I did shift from being against the death penalty to thinking that if it has a significant deterrent effect it’s probably justified.” The quote is from Cass Sunstein, one of the leading liberal law professors.

But there's always another on-the-other-hand: "A single capital litigation can cost more than $1 million. It is at least possible that devoting that money to crime prevention would prevent more murders than whatever number, if any, an execution would deter." That's a testable proposition, I would think.

The article also leaves undiscussed the possibility that the death penalty reinforces the high level of violence in American culture, something that could explain why death-penalty-using America is more violent than Europe, and why the American South is more violent than elsewhere. The problem with this argument is that it's very squishy and not easily testable. I suppose that doesn't make it impossible though.

The trickiest question, assuming that spending money on policing won't work, is what to do with the current, less than perfectly settled state of knowledge. I'd think that applying the precautionary principle suggests that the most cautious approach from a saving lives perspective is to support the death penalty. Not something I'm quite ready to do, though.

And one more wrinkle: I usually think it obvious that policies proven to work should be expanded. If the death penalty works, are we making a mistake in the reduced number of executions over the last few years?