Does Capital Punishment Deter Crime? - Island of Sanity

Law

Does Capital Punishment Deter Crime?

Advocates of capital punishment routinely argue that statistics prove that it deters crime. Opponents of capital punishment just as routinely argue that statistics prove that it does not.

I suppose a naive person might find this disagreement puzzling. Even if we cannot agree on moral questions, surely we could at least agree on basic facts. I mean, it would be understandable if an anti-capital punishment person said that, yes, it does deter crime, but it is still wrong because it is cruel and barbaric; or if a pro-capital punishment said, okay, it doesn't deter crime any more than life imprisonment or some other punishment, but it is still right because it is just. But can't we at least agree on the underlying facts?

But as I'm sure we're all aware these days, you can twist statistics to prove almost anything. Statisticians have developed many sophisticated techniques to carefully analyze data. People with a point to prove can abuse these techniques to distort the data.

But I'm a simple guy, so I decided to look at the simple statistics. Let's just look at the raw numbers: no clever analysis, no involved mathematical manipulation, just look at the numbers.

So, using statistics from the United States Department of Justice website, here's my graph number 1: The homicide rate for each year since 1950.1 The rate is given as the number of homicides for every million people.

Graph 1: Homicide Rate

A casual glance at this graph clearly shows that homicide rates increased sharply beginning about 1965 or 1966, they took a steep dive from 1980 to 1985, started back up again until 1991 - 1992, and now appear to be inching down.

Surely a reasonable, concerned person could ask if there is any apparent cause for the sudden sharp increase in the late 60's. And surely we could look with hope at the drop in the early 80's, and ask if there was not something that was happening then that we could reproduce.

So let's look at another graph. Graph number 2 shows the homicide rate, just as above, and on top of this I show the number of cases where capital punishment was imposed.2

Graph 2: Homicide Rate vs Executions

Note the interesting correlations. The number of executions plummeted from 47 in 1962 to 2 in 1967 to zero in 1968. The homicide rate, which had been holding steady around 50 throughout the 50's, started up in 1965, just two years after executions began their plummet. The biggest increase in one year came in 1967, the same year that the last person was executed.

So okay, maybe this was simply a coincidence. Capital punishment was reinstated a decade later. What happened then?

In 1976 the Supreme Court issued several decisions in which they basically backtracked and again allowed capital punishment. (They didn't quite say that they were changing their minds or admitting error, but rather that the flaws which they had discovered in the previous capital punishment laws had now been corrected.) The first person was actually executed in 1977. In the very year of these Supreme Court decisions, the homicide rate plummeted. But no more than two people were actually executed in any one year through 1982, and so perhaps criminals concluded that the danger of execution was remote, and the homicide rate crawled back up. Then the number of executions suddenly went up in 1983, and in that year the homicide rate showed its biggest one-year drop. With the sudden surge in executions in 1996, the homicide rate again fell.

Indeed, just looking at this graph we can see that the homicide rate is almost the mirror image of the number of executions. Consistently when the number of executions goes down, the homicide rate goes up, and when the number of executions goes up, the homicide rate goes down. The only major exception to this is the fall in homicides in 1976, which came before executions re-started. But this is easily explainable by the fact that the court decisions allowing executions to resume came a year or two before executions actually did resume. Criminals may have been responding to press reports that capital punishment was once again going to take place, in advance of it actually happening.

I'm sure that opponents of capital punishment will say that my analysis here is too simplistic; that I have failed to take other factors into account; that this correlation between execution rates and homicide rates is pure coincidence, and that other factors explain why homicide rates went up and down at these times that had nothing to do with the number of executions.

To which I reply, Well, maybe, but I think you have an awfully hard sell. If there was just one point of correlation, it might be explained by coincidence. That is, if the homicide rate had gone up when capital punishment was abolished, but when capital punishment was re-instated the homicide rate had remained unchanged, or had gone up further, one might reasonably say that the first correlation was simply coincidence. But when we can clearly see that the two numbers mirror each other, consistently over a period of almost fifty years, attributing this to coincidence gets pretty hard to believe.

The obvious conclusion from looking at the statistics, without any fancy "analysis" or "factoring out of other factors", is that capital punishment does deter murder.

Footnotes

1. Source: United States Department of Justice, Bureau of Justice Statistics. "Homicide rates from the Vital Statistics." http://www.ojp.usdoj.gov/bjs/glance/hmrt.txt (March 1998). They in turn give the source of their data as the "National Center for Health Statistics, Vital Statistics". They give the homicide rate per 100,000 people -- I have adjusted this to "per million".

Afterword

Lest I be accused of the same offense of which I accuse others, I will here freely discuss how I selected the data which I present.

I used homicide rate, i.e. homicides per million people, rather than the simple number of homicides, for two reasons. One, the data from the Department of Justice presented the numbers that way, so it was easiest to just take their numbers as given rather than looking up the population of the country each year and computing the simple number. (And I also would have run into the problem of possibly using different population statistics than the Department of Justice did, and thus introducing errors into the data.) Second, this is a good way to present the data anyway as it reflects the likelihood that any one person will decide to commit murder, which is the question under consideration. A graph of the simple number of murders would presumably show an overall upward trend reflecting growing population, and thus masking any deterrence or lack of deterrence.

On the other hand I used the simple number of executions rather than any "execution rate". Arguably it would be more accurate to plot executions as a percentage of the number of homicides, to show the probability that any given murderer would, in fact, be executed. I didn't do this for the simple reason that the Department of Justice figures did not present the numbers this way, and to compute them I would have had to convert homicide rate to number of homicides, and then compare this number to number of executions. This would have involved a lot of calculation using statistics from other sources, and so would not only have been a lot of work but could have introduced errors as mentioned above.

I converted homicide rates from homicides per 100,000, as the Department of Justice tabulates them, to homicides per million, simply to allow me to plot both homicide rates and number of executions on the same scale. That is, this conversion put the homicide rates in the 50 to 100 range or so, and executions were in the 0 to 100 range, so this conversion made it possible to plot both on the same graph without one being reduced to a line barely distinguishable from zero.

A more serious criticism could be leveled at my selection of the time range to include. I could not find any data on executions before 1930 or after 1996, so that was the widest possible range I could use. Clearly I had to go before the mid-60's to get before-and-after abolition comparisons, so going back to 1950 seemed about the shortest range to give meaningful numbers for comparison, and I went through the latest numbers because we surely want to see where things are right now.

Lest you wonder, analysis of earlier numbers reveals that the homicide rate went up sharply in the 20's and early 30's and fell in the late 30's and 40's. As I don't have data on executions before 1930, I cannot say if there is any correlation between the execution rate then and this increase. There were increasing numbers of executions in the late 30's as the homicide rate was coming down, but the correlation is nowhere near as obvious or dramatic as it is for the time period I discuss in the body of this article. Thus, I conclude that the movement of homicide rates in the 30's probably cannot be entirely explained by any deterrent effect of capital punishment, but must be at least partly attributed to other factors. I do not see this as in any way a refutation of the point of this article: neither I nor any one else I have ever heard is claiming that the deterrent effect of capital punishment is the only thing which affects homicide rates, merely that it is an import factor. No one denies that there could be other factors.

If anyone wants to accuse me of "cooking the data" to bias the results, based on the above or other considerations, that is your privilege. But I think in fairness you must show some equally simple, defensible selection of data that shows no correlation between capital punishment and murder rates. Otherwise, I think you can fairly be accused of saying simply, "That data must be invalid because it does not agree with what I just know must be true. No further analysis is necessary."