The study examined the relationship between the number of executions and the number of murders in the U.S. for the 26-year period from 1979 to 2004, using data from publicly available FBI sources. The chart nearby shows the number of executions and murders by year. There seems to be an obvious negative correlation in that when executions increase, murders decrease, and when executions decrease, murders increase.
In the early 1980s, the return of the death penalty was associated with a drop in the number of murders. In the mid-to-late 1980s, when the number of executions stabilized at about 20 per year, the number of murders increased. Throughout the 1990s, our society increased the number of executions, and the number of murders plummeted. Since 2001, there has been a decline in executions and an increase in murders.
It is possible that this correlated relationship could be mere coincidence, so we did a regression analysis on the 26-year relationship. The association was significant at the .00005 level, which meant the odds against the pattern being simply a random happening are about 18,000 to one. Further analysis revealed that each execution seems to be associated with 71 fewer murders in the year the execution took place.
While it is clear that the number of murders is inversely correlated to the number of executions, it is dangerous to infer causal relationships through correlative data. Causation can be a two-way street, but not in the case of capital punishment. It may be logical that more executions could lead to fewer murders, but it is not at all logical that fewer murders could cause more executions.
A second difficulty with strong correlative data is that of timing. Causes should come before effects, so we correlated each year's executions to the following year's murders and found the results to be even more dramatic. The association was significant at the .00003 level, which meant the odds against the random happening are longer than 34,000 to one. Each execution was associated with 74 fewer murders the following year.
http://online.wsj.co...397079767680173