Yahoo! news has cheerfully reported that Texas executes more prisoners than any other state; it attributes this fact to the culture of the state:
“In Texas you have all the elements lined up. Public support, a governor that supports it and supportive courts,” said Richard Dieter, executive director of the Death Penalty Information Center.
Mr. Dieter ignores the fact that Texas has 23 million people (or 7.7% of the US population); it’s above-average size surely accounts for its above-average number of executions. Fourteen states and DC do not have the death penalty; of the states which do, only California has a larger population than Texas. Furthermore, in 2003, the governor of Illinois commuted all of the death sentences of the 156 inmates in the state. Of the roughly 300 million people in the United States, only 229 million of those live in states which have the death penalty. The population of Texas represents 10.3% of those persons.
Dallas and Houston are both among the top 10 highest-murder rate cities in the US and Europe. Texas has executed 398 persons since the death penalty was reintroduced in the state in 1982, which represents 36.5% of the total executions. Now, this is still higher than one would expect from a state which has about a tenth of the population, but its execution rate per capita (1.69 executions per 100,000 residents) is nearly identical to that of Delaware (1.64 executions per 100,000 residents) and much lower than that of Oklahoma (2.37 executions per 100,000 residents). Yet, newspaper articles do not decry Oklahoma as the Death State; only Texas is maligned.
Note: executions have occurred since 1976; population figures are taken from the 2000 Census. Executions per 100,000 residents is noted as total executions since 1976 divided by 1/100,000th of the current population.
The New York Times also ran amok with its statistics. It reported that either women or men lie on anonymous sex surveys. Its evidence? Women report a median of four partners in their lifetime, while men report a median of seven. As this is heterosexual intercourse, every sex partner that a man has will result in a woman having one as well. One possible interpretation is that people lie; several others are also valid. A summary:
- Median v. mean: if a small number of women have sex with a lot of men, the arithmetic mean will be the same, but the medians will be different. A different median only tells us that the two bell curves have different shapes.
- The #men=#women hypothesis only holds true in a closed system (whereby the only partners men have had are the women in the sample, and vice versa), or where the sample size is sufficiently large so as to negate any effects of leaving out persons. It is quite possible that a survey left out especially chaste men or promiscuous women.
- Even in a closed system, the numbers of persons must be equal in order for the arithmetic mean to be identical. Consider a polygamous family who never had extramarital sex. A man with four wives will report four sexual partners over his lifetime, while the women will each report one. Here, the mean and median of sexual experiences among the women is 1; the mean and median for men is 4.
- Outliers. Often, statisticians will remove outliers in a data set; however, outliers (rare men who have not slept with anyone, or the rare woman with 100 partners) explain the discrepancy quite well.
Of course, it is possible that men overstate their conquests and women understate the number of partners they had; however, it is also likely that the different distributions of the respective bell curves and sampling issues have skewed the results. A person could survey, for example, a class of college seniors and ask them how many partners have they had, of the opposite sex, who are also seniors at that college. Such a survey, if taken of the entire senior class, would eliminate the problem of counting only 1/2 of a pairing. If there are different numbers of men and women, the study authors would have to content themselves with using totals and not averages, as counterintuitive as it is. The results of that study could more conclusively determine if the noted discrepancies are due to reporting or different distributions of behaviour.