<div>Aha.</div>
<div> </div>
<div>I think I understand, having read both of the main papers cited. There are two topics of discussion here:<br><br>(1) The Naci Moran study, which was conducted on the 1977-1997 dataset.</div>
<div> </div>
<div>and <br><br>(2) An Emory University metastudy of other death penalty studies conducted on the 1977-1997 dataset.</div>
<div><br>There are several exceptional factors in this dataset in particular. First, that between 1972 and 1976, all death sentences were commuted by the Supreme Court's decision that the death penalty was, in all cases, unconstitutional. Second, that between 1977 and 1997, there was a sharp general decline in the crime rate, including murder. Third, that Texas performed just over one third of all executions in the United States.
</div>
<div><br>This introduces substantial distortions into any study of the data. </div>
<div> </div>
<div>First, in January 1977, the only Americans currently on Death Row were Americans convicted in late 1976. The death penalty can take close to a decade to execute -- you're looking at most death penalties in 1976 actually being executed in the early 1980s. For this reason, this particular dataset contains exactly zero executions over the period with the greatest crime rate, with a slow increase (due to the lag in execution) over the duration of the dataset. As there are no claims made that the death penalty has a general deterrent effect on crimes not punishable by death, then it would seem that there are two synchronous but not correlated events here: an increase in executions following a favorable Supreme Court decision, and a decrease in the general crime rate.
</div>
<div> </div>
<div>Second, death penalties -- even in the case of murder -- constitute a very small fraction of the total sentences handed down. Examining the death penalty without regard to variables correlated to the likelihood of a death penalty being executed ignores the deterrent effects of changes in sentencing. It is not unreasonable, for instance, to believe that a state's willingness to execute the death penalty is correlated with the severity of its sentencing guidelines for murder.
</div>
<div> </div>
<div>Third, Texas. 1/3 of the executions from that dataset occurred in Texas. Variations in Texas policy, Texas demographics, and Texas sentencing guidelines exercise an effect truly disproportionate to the population of the state.
</div>
<div> </div>
<div>If you want to read about the methodological problems with drawing conclusions about murder in particular based on that dataset, you can read (essentially) the article I cribbed my response from here:</div>
<div> </div>
<div><a href="http://repositories.cdlib.org/cgi/viewcontent.cgi?article=1059&context=uclastat">http://repositories.cdlib.org/cgi/viewcontent.cgi?article=1059&context=uclastat</a></div>
<div> </div>
<div>-- ACS</div>