Ever since Paul Ehrlich’s 1975 paper on the topic, the consensus amongst economists has been that the death penalty is a powerful deterrent, with studies finding that for each execution anywhere from a couple to dozens of lives are saved from the deterrent power that the death penalty exerts.
This has provided tremendous empirical support for the viability of the death penalty. Over at the Becker-Posner blog, Gary Becker went so far as to say that:
I support the use of capital punishment for persons convicted of murder because, and only because, I believe it deters murders. If I did not believe that, I would be opposed because revenge and the other possible motives that are mentioned and discussed by Posner, should not be a basis for public policy.
(Cass Sunstein and Adrian Vermeule take a similar line of reasoning here.) With that in mind, John Donohue (of Levitt-Donohue “Abortion paper” fame) gave a talk at a workshop at the University of Chicago Law School on his new paper that takes a closer look at past econometric studies of the death penalty’s deterrent power. Interestingly enough Donohue finds in this paper that past studies have, for the most part, been wrong (another interesting result brought to light in the talk was that the shittiness of prison conditions is apparently an overwhelmingly powerful deterrent, at least according to Katz, Levitt, and Shustorovich).
For a really lengthy discussion on the viability of different methods to tease out the causal relationship between the death penalty and the rate of homicide, read the paper. I’d talk more about that, but I figure no one would actually read it. The aspect of the talk that was much more bloggable (aside awesomeness that was the back and forth between Judge Posner, Cass Sunstein, Richard Epstein, etc. and Donohue) was what Donohue had to say about econometric studies.
What Donohue found was that in replicating past studies, changing some of the simplest parameters of the models caused the results to completely reverse (to the extent that results found that each execution actually caused more people to be murdered). One aspect of Donohue’s was against the use of such sensitive results in emotionally charged debates, like the deterrent power of the death penalty or the impact on crime of guns. Such empirical research is extremely powerful, especially when it can be simplified into: X executions prevents Y murders.
But as a result of the data’s sensitivity, the biases of authors can easily creep into the way one parameter is defined, which in turn can define the causal result the paper ends up proposing. This sort of controversy is what led to the whole Levitt-Lott bruhaha on whether Lott manipulated his data in order to get a result on guns reducing crime that he wanted (as a result Levitt has found himself serviced with a lawsuit for defamation, but that’s neither here nor there).
Also, critical in Donohue’s talk was the significance that these results play in public policy debates. All different ideologies are apt to pick up and run with any empirical result they find on the web, and as a result, it seems that economists ought to be significantly more conservative with their conclusions then they have been, especially on extremely political issues.