Saturday, April 26, 2014

Death Penalty Deterrence and Panel Studies

Last Thursday I went to the annual banquet of the Pittsburgh Chapter.  The keynote speaker was Daniel Nagin from Carnegie Mellon University.  He was on a National Research Council panel assessing the deterrence effect of the death penalty on the homicide rate in the US.  He described two different types of studies used to assess this effect: 1) panel studies which use correlation and regression methods on state homicide and execution rates to establish an association and 2) time series studies which look at how the homicide rate changes in a state after there is an execution or a change in the law.  In an article in the April edition of the American Statistical Association's magazine Significance,

Nagin argues that past studies on this topic using these methods are inconsistent.  He says they do not take into account periodic fluctuations in the homicide rate and the number of executions.  Also they do not take into account whether potential murderers actually consider that they might have to face the death penalty.  In the case of white supremacist Frazier Glenn Miller who killed three people outside of Jewish centers in Overland, KS, it seems clear that he was not thinking of the possibility of execution as Kansas has the death penalty (though it hasn't executed anyone since 1976, while neighboring state Missouri, where he lived has executed 76 so maybe he was thinking of deterrence).  There are many other cases of murder in death penalty states that could be cited as evidence of either a lack or a presence of deterrence. The problem is adequately measuring and tabulating this effect, for every murderer that was not deterred how many non-murderers were deterred?

The other reason I write of this speech is that I have frequently used the panel methodology on this blog.  Most recently, I've posted on the concentration of hate groups in the US and on it's correlation with health outcomes.   Nagin's criticisms of panel studies could possibly be applied to posts like these.  This is the first post I have done on a statistic as changeable over time as the death penalty.  The statistics I have compiled for this blog are more static, partly because they are only made available either annually, quarterly or monthly.  Also statistics like uninsured rates, per capita income, etc. can be more static from year to year. The point of those posts is to find meaningful patterns in our world.  Establishing cause and effect is a lot more difficult which Nagin is trying to prove.

**Related Posts**

Income and Life Expectancy. What does it Tell Us About US?