The New York Times publishes a misleading claim of racial bias in the death penalty

[The following post originally appeared here on April 30, 2008.]

Yesterday's New York Times had an article about a study of the death penalty in Harris County, Texas. The article began as follows:

About 1,100 people have been executed in the United States in the last three decades. Harris County, Tex., which includes Houston, accounts for more than 100 of those executions. Indeed, Harris County has sent more people to the death chamber than any state but Texas itself.

Yet Harris County’s capital justice system has not been the subject of intensive research — until now. A new study to be published in The Houston Law Review this fall has found two sorts of racial disparities in the administration of the death penalty there, one commonplace and one surprising

The unexceptional finding is that defendants who kill whites are more likely to be sentenced to death than those who kill blacks. More than 20 studies around the nation have come to similar conclusions.

But the new study also detected a more straightforward disparity. It found that the race of the defendant by itself plays a major role in explaining who is sentenced to death.

The last sentence (which I have highlighted in bold) is the main point of contention. The article eventually quotes someone who is suspicious of the study's methodology. But a reader would have to go deep into the article to see that quote. And the article fails to point out the worst flaws in the study.

To see what is really going on, you need to look at the study itself. You can find it here. A few facts jump out:

First of all, as the article notes, of 100 defendants indicted for capital crimes in Harris County, 27% of the blacks and 25% of the whites were sentenced to death. No bias there.

However, after controlling for the mitigating and aggravating factors such as the heinousness of the crime, the study found bias.

But the study measured the mitigating and aggravating factors by looking at newspaper coverage of the case! So what if the newspaper coverage depends on the assailant's race? Many people have complained that a murder in a white area is big news, while a murder in a black area is not. So if the white area murderers get more thorough newspaper coverage than the black murderers, the study is going to find more aggravating factors, which makes the white murderers appear "worse" than the black ones, which in turn makes it appear unfair that they are sentenced to death at an equal rate.

That's not all. The study also reports that there were plea bargains in 38% of the cases with white defendants, and only 28% of the cases with black defendants. In capital cases, it is common for the defendant to agree to plead guilty if the prosecutor is willing to take the death penalty off the table. The statistics suggest (but do not prove) that the white defendants are more willing to plea bargain. Now, all other things being equal, a group of defendants that is more willing to plea bargain should be sentenced to death less often. But this is a result of the decisions of the defendant's themselves.

So you have a study which started with a situation that looks raceially neutral (27% of black capital defendants sentenced to death vs. 25% of whites). There are several confounding factors. Based on the newspaper coverage, the heinousness of the murders committed by whites appears to be worse than the heinousness of the murders committed by blacks. So the study controls for heinousness. Based on plea bargaining data, whites appear to be more willing to plea bargain than blacks. But the study's author, who realizes that a conclusion of racial bias will get more press than a conclusion of no racial bias, elects not to control for the defendant's willingness to plea bargain. And lo and behold, the study ends up concluding that there is racial bias.

The study's author is rewarded with coverage in the New York Times.