Tuesday, 1 July 2008

Wolfers on the death penalty

At Freakonomics Justin Wolfers blogs on How the Supreme Court Misread My Research: Empirics and the Death Penalty. Wolfers writes
Cass Sunstein and I have an Op-Ed in today’s Washington Post, discussing the mis-reading of available empirical evidence in recent death penalty jurisprudence.

Some background: A recent Supreme Court ruling (Baze v. Rees) reaffirmed the constitutionality of the death penalty, and along the way, the justices revisited the empirical literature on whether the death penalty has been shown to be an effective deterrent.

Justice Stevens cited research by John Donohue and myself in concluding that “there remains no reliable statistical evidence that capital punishment in fact deters potential offenders.” Countering this, Justice Scalia cited a paper by Cass Sunstein and Adrian Vermeule, in arguing that the data do in fact point to deterrence.

As two of the supposed flag bearers for the competing views cited by the court, Sunstein and I thought it worth poring over the data to see what we agree on. It turns out there’s a lot of agreement between us:

“In short, the best reading of the accumulated data is that they do not establish a deterrent effect of the death penalty.”

He adds some useful links to the debate on the death penalty for those interested.
The full Op-Ed is available here (or else ungated, here). A readable introduction to this literature is available here.

For those interested in the empirical observations in our Op-Ed, I have put together a longer document discussing that evidence, available here; there are plenty of pretty pictures and pointers to background material.

Don’t want to take my word on the evidence? I have collected the available data and posted it here — feel free to generate your own findings. And read Levitt’s $0.02, here, here, and here.
Greg Mankiw points out that the death penalty does solve the problem of recidivism, while Ronald Bailey at Reason magazine asks Does It Really Matter If The Death Penalty Deters Murderers? In the article Bailey argues
The reason to retain the death penalty is vengeance, or as more polite people put it, retribution.
But notes at the end of the article
... in the era of post-conviction exoneration through DNA testing, I have become much more uncomfortable about the possibility that an innocent person might be executed.

1 comment:

dudleysharp said...

There are some problems with:

A Death Penalty Puzzle: The Murky Evidence for and Against Deterrence (Washington Post, By Cass R. Sunstein and Justin Wolfers, June 30, 2008; Page A11)

The article states: "But that suggestion (of the many recent studies finding for deterrence) actually catalyzed Donohue and Wolfers's study of available empirical evidence. Existing studies contain significant statistical errors, and slightly different approaches yield widely varying findings, a problem exacerbated by researchers' tendency to report only those results supporting their conclusions."

REPLY: ". . . researchers' tendency to report only those results supporting their conclusions."! My goodness, this seems much like the pot calling the kettle black. Please see the four replies to Donahue and Wolfers, at bottom - replies which, collectively, appear to lay waste to Donahue and Wolfers criticism. It should also be pointed out that Donahue and Wolfers chose not to publish their criticism in a peer reviewed publication.

The article continues: "In short, the best reading of the accumulated data is that they do not establish a deterrent effect of the death penalty."

REPLY: I think it is arguable that is an unfair reading of the data.

My take is that it would be more accurate to state that the social science of determining the measurable effect of general deterrence is challenging, at best. The strength of the recent econometric studies is suggestive of a deterrent effect of the death penalty and the criticism of those studies seems much less robust than the studies, themselves.

In general terms, the weight of the evidence is that all prospects for a negative outcome deter some. There appears to be few, if any, exceptions to that rule.

The article continues: "But as executions resume, the debates over the death penalty should not be distorted by a misunderstanding of what the evidence actually shows."

REPLY: Again, it appears that misunderstanding is, precisely, what this newest article was distributing, however unintended.

The article continues: "A prominent line of reasoning, endorsed by several justices, holds that if capital punishment fails to deter crime, it serves no useful purpose and hence is cruel and unusual, violating the Eighth Amendment." "While some favor the death penalty on retributive grounds, many others (including President Bush) argue that the only sound reason for capital punishment is to deter murder."

REPLY: The lack of balance or understanding by S&W appears quite robust. There are many well defined reasons for criminal sanction, outside of deterrence. Retribution, justice, upholding the social contract, incapacitation, among others. Obviously, capital punishment does not violate the Eight Amendment, as both the 5th and 14th Amendments assert the constitutionality of the death penalty. These are well known and important facts, not in S&W's article.

Both the current Pres. Bush and former VP Al Gore stated they support the death penalty because of deterrence. However, it is important to clarify. Does anyone believe that either Gore of Bush would support the execution of someone if it wasn't deserved? I suggest the answer is no. It's a fair question. Ask them. Then ask them, if the death penalty did not deter like men as Bib Laden or Hitler, would you still support their executions? Does anyone not know their answer? Of course not.

The article continues: "We concur with Scalia that if a strong deterrent effect could be demonstrated, a plausible argument could be made on behalf of executions. But what if the evidence is inconclusive?"

REPLY: Inconclusive? The weight of the evidence is that some are deterred by any prospect of a negative outcome. But, let's project an imaginary world where the evidence is completely neutral on the effects of negative prospects, where there is no evidence of what incentives mean to behavior.

We have two equally balanced prospects. The death penalty/executions deter and the death penalty/executions don't deter.

This prospect is neither inconclusive or equally balanced, because you have a prospect between sparing innocent life, via death penalty/execution deterrence or a prospect of death penalty/execution, with no deterrence, but enhanced incapacitation.

If deterrence is inconclusive, the prospect of saving innocent lives is not.

In addition, the deterrence calculus becomes- we execute and save additional innocent lives (via deterrence) or we execute and don't save additional innocent lives (via deterrence). The weight falls upon saving additional lives, if we have inconclusive data, because the weight is greater for saving additional innocent lives than on not saving additional innocent lives.

With this type of calculus, the only argument for not executing is if the evidence is clear that the death penalty/execution is sacrificing more innocents, via brutalization - the prospect that by executing, you are inspiring potential murderers to become real murderers. Brutalization is discounted because the evidence for deterrence is much more robust and the practical evidence that crimes will increase because of raised sanction doesn't exist.

As noted, there are a number of good reasons to argue in favor of the death penalty, other than deterrence. Even if the death penalty is a very marginal deterrent, it is still a robust argument in favor of the sanction.

Let's say that every four executions deters one potential murderer and every twenty death sentences deters one potential murderer. That would mean that around 650 innocents had been spared since 1973. Even marginal deterrence equals major benefit.

This does not take into account the two other ways that the death penalty saves additional innocent lives. Enhanced incapacitation and enhanced due process.

Sincerely, Dudley Sharp


four replies to Donahue and Wolfers

Hashem Dezhbakhsh & Paul H. Rubin
From the 'Econometrics of Capital Punishment' to the 'Capital Punishment' of Econometrics: On the Use and Abuse of Sensitivity Analysis (September 2007). Emory Law and Economics Research Paper No. 07-18, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1018533

Abstract: The academic debate over the deterrent effect of capital punishment has intensified again with a major policy outcome at stake. About two dozen empirical studies have recently emerged that explore the issue. Donohue and Wolfers (2005) claim to have examined the recent studies and shown the evidence is not robust to specification changes. We argue that the narrow scope of their study does not warrant this claim.


(2006) "This analysis shows that ((Donohue and Wolfers' "D&W's")attempts to make the deterrence effect disappear are ineffective." (p 16)
--- Existence of the death penalty, in law, has a statistically significant impact on reducing murders. (p 23)
--- Execution rates show significant impact in reducing murders. (p 13 & 23)
--- Death row commutations, and other removals, increase murders. (p13 & 23)
--- The criticism of our studies is flawed and does not effect the strength of the measured deterrent effect.
"The Impact of Incentives On Human Behavior: Can we Make It Disappear? The Case of the Death Penalty", Naci H. Mocan, R. Kaj Grittings, NBER Working Paper, 10/06, www(dot)nber.org/papers/w12631

(2006) " . . . (Donohue and Wolfers' "D&W") criticisms of Zimmerman's analysis are misrepresentative, moot or unsupportable in terms of the analyses they perform." "It is shown that Zimmerman's published empirical results, or the conclusions drawn from them, are not in any way refuted by D&W's critique." (pg 3) "This later estimate suggests that each execution deters 14 murders on average . . .". (pg 7) "It is shown that D&W made a number of serious misinterpretations in their review of Zimmerman's study and that none of the analyses put forward by D&W (which ostensibly refute Zimmerman's original results and conclusions) hold up under scrutiny. (pg8) " . . . D&W do not even report Zimmerman's "preferred" results correctly, and then proceed by carrying on this error throughout the remainder of their critique."(pg8) "Of course, (D&W's) omission tends to create a strong impression that Zimmerman's analysis 'purports to find reliable relationships between executions and homicides', when his actual conclusions regarding the deterrent effect of capital punishment are far more agnostic." (pg10) " . . . D&W's method of interpreting their results is not consistent with that proscribed by the received econometric literature on randomized testing . . .". "As such, D&W's interpretation of their randomized test in itself does not (and cannot) reasonably lead one to conclude that Zimmerman's estimates suggesting a deterrent effect of capital punishment are spurious." (pg12) " . . . D&W do not appear to have interpreted their randomization test in any meaningful fashion." (pg14) " . . . the state clustering correction employed by D&W may not be producing statistically meaningful results." (pg16) "And while D&W once lamented that recent econometric studies purporting to demonstrate a deterrent effect of capital punishment yield 'heat rather than light', as shown herein, their criticisms of Zimmerman (2004) tend to yield 'smoke rather than fire'."(pg26)
Zimmerman, Paul R., "On the Uses and 'Abuses' of Empirical Evidence in
the Death Penalty Debate" (November 2006). ssrn(dot)com/abstract=948424


(2007) "Had (D&W's) paper been subjected to the normal blind peer review process in an authoritative economic journal it is highly unlikely that it would have survived intact , if at all. "

"(D&W's) Quibbling over numerous and sometimes meaningless statistical issues obscures the picture painted by the cumulative effect of the nearly dozen studies published since the turn of the 21st century."

"Using differing methodologies and data sets at least five groups of scholars each working independently (and often without knowledge of the others) have arrived at the same conclusion—there is significant and robust evidence that executions deter some homicides. While there may be merit in some of (D&W's) specific criticisms, none addresses the totality of the collection of studies. The probability that chance alone explains the coincidence of these virtually simultaneous conclusions is negligible."

"DW’s unsupported claim that the appropriate variable in studies of deterrence using these borrowed tools from portfolio analysis is the amount or level of homicides in the respective jurisdictions. This claim is without theoretical basis or empirical precedent. "

"With regard to DW’s specific comments on our two papers (Cloninger & Marchesini, 2001 & 2006) we find very little requiring defense. Implicit in their critique, and explicitly stated in private communications, DW were able to replicate our results based on data we furnished, at their request, as well as data they acquired independently. "

"Reflections on a Critique", Dale O Cloninger and Roberto Marchesini, forthcoming Applied Economic Letters