Death Penalty again

Via GrrlScientist I find the latest story claiming:

What gets little notice, however, is a series of academic studies over the last half-dozen years that claim to settle a once hotly debated argument–whether the death penalty acts as a deterrent to murder. The analyses say yes. They count between three and 18 lives that would be saved by the execution of each convicted killer.

However, there is little new since the last time I wrote about Donohue and Wolfers’ take down of those studies. They found that the results of those studies were not robust — simple changes to the models made the effect go away. Here’s an interesting graph from Donohue and Wolfers:


It shows the size of the effect (vertical axis) versus its standard error. The interesting thing is that there is a strong relation between these two numbers. Studies that have large standard errors also find a large effect. Which is just as well, because if they didn’t find a large effect, the result would not be statistically significant. This is evidence for publication bias. Even if there is no effect, by chance some specifications will be significant at the 5% level and these are the ones that get published. In this case, the published results will tend to be barely significant, and the graph shows that as well.

Steven Levitt comments:

Given the evidence I’ve examined, I believe that Wolfers is on the right side of this debate. There are recent studies of the death penalty — most bad, but some reasonable — that find it has a deterrent effect on crime. Wolfers and John Donohue published an article in the Stanford Law Review two years ago that decimated most of the research on the subject.

Analyses of data that stretches further back in time, when there were many more executions and thus more opportunities to test the hypothesis, are far less charitable to death penalty advocates. On top of that, as we wrote in Freakonomics, if you do back-of-the-envelope calculations, it becomes clear that no rational criminal should be deterred by the death penalty, since the punishment is too distant and too unlikely to merit much attention. As such, economists who argue that the death penalty works are put in the uncomfortable position of having to argue that criminals are irrationally overreacting when they are deterred by it.

Justin Wolfers has more links to the back and forth arguments, while the latest from the other side is this paper from Mocan and Gittings.


  1. #1 Jonathan Arnold
    June 11, 2007

    Today’s Boston Globe featured this story on page 2. Combined with the ridiculous Discovery Institute op-ed on the Creationism/Evolution “debate”, the Globe sunk to new lows in publishing junk science, spurring my first ever letter to the editor.

    Just reading the story would make anyone’s BS detector ring off the charts. Hard coded numbers like “every execution saves 18 lives” are just too extremely detailed to be believed. Plug some numbers in and out pops an 18 – patently outrageous.

  2. #2 BWV
    June 11, 2007

    I agree that the statistical model is bunk – they seem the rely on the assumption that one can identify every other explanatory variable in changes in the murder rate and then assign the residual to capital punishment.

    Logically, capital punishment must have some deterrence effect there is some portion of the potential criminal population that will be deterred by it without creating any offsetting incentives by any other portion of the population.

    But, the deterrence argument is moot from a moral standpoint. For example crucifying murderers on national TV would be a likely deterrent

  3. #3 z
    June 11, 2007

    “But, the deterrence argument is moot from a moral standpoint. For example crucifying murderers on national TV would be a likely deterrent”

    The whole moral calculus, i.e. if you save more innocent lives than you conceivably lose it’s a good thing, is flawed; because it has an inexorable corollary that you should therefore frame and execute an innocent person ASAP when the real culprit can’t be caught, in order to maintain the deterrent effect. Such things happen in totalitarian countries for such reason. Shows what happens when you apply Cartesian reduction to irreducible things.

  4. #4 Ian Gould
    June 12, 2007


    This was the standard approch in Impeial China, if no obvious suspects present themselves torture the nearest servant or peasant until they confess.

  5. #5 Brian Weatherson
    June 14, 2007

    I agree with Z that it would be horrible to frame innocent people to maintain the deterrence effect. But in what possible way is this *Cartesian*? It certainly doesn’t trace to anything in Descartes. (I’m no expert on Descartes, but I can say this with some confidence.) And I don’t even know how you could start with anything Descartes wrote and end up at that position. Certainly the _Meditations_ doesn’t push that way – it’s anti-reductionist if anything – but maybe there are parts of the _Methods_ (which I don’t know well) you could spin in a reductionist way. Or something else?

    I’m sure I’m missing something here, since I have very little knowledge of the myriad ways people have used and abused Descartes, but this one strikes me as particularly bizarre.

  6. #6 z
    June 14, 2007

    Well, I may well be using this way off the mark. But way back there when I had time to study stuff that wouldn’t earn me a living, I realized how the West had bought wholesale into Cartesian reductionism, i.e. things could be studied by decomposing them into their component parts, studying the individual behavior of those parts, and studying how they interacted when assembled. Science, obviously, but even art; a canvas full of nothing but red paint is the equivalent of a flask full of nothing but lactic acid dehydrogenase.

    And it’s worked out well for the West, by and large; but it occurred to me how many things don’t fall into that, and are irreducible to analysis. Like love; I don’t fall in love with somebody because of 10 points for a cute face and seven points for liking dogs and 3 points for sturdy feet, it’s irreducible all or nothing. And I thought it applied here; you can’t judge the morality of the death penalty by subtracting the innocent lives estimated to be saved from the innocent lives estimated to be lost and come out with a value judgement.

    As I said, I’m not a philosopher, it’s the only course I ever got a C in, so I could be totally off the wall. In which case, I will definitely go to Wikipedia and demand my point of view be represented.