The Colours of Death Row
If you get involved in a trial for the murder of a white man in the US, you'd better be white too, especially when the court is in a Southern state. According to a new working paper by Alberto Alesina (Harvard University and IGIER) and Eliana La Ferrara (Department of Economics and IGIER) (A Test of Racial Bias in Capital Sentencing, IGIER Working Paper n. 387), first degree American courts make do with a lower standard of proof, when the victim is white, if the defendant belongs to a racial minority. Lower courts, in other words, are racially biased against minority defendants suspected of having killed white men. And the trend is driven by Southern states.
Racial bias is one of the traditional arguments against the death penalty in the US, but it is simplistically argued with statements like "African Americans are disproportionately represented among people condemned to death" – which could simply mean that crime rates are higher among minorities, or that they can afford less experienced lawyers. Alesina and La Ferrara propose a rigorous test of racial bias based on the distribution of judicial errors by first degree courts, i.e. cases in which a death penalty imposed by a lower court is reversed by higher courts. The key insight is that, if the court is unbiased, the likelihood to making an error when imposing a death sentence should be the same regardless of whether the victim is white or not. Crucial to the validity of their test is the assumption that, while lower courts can be biased, higher ones are not – or at least are biased to a lesser degree. "What we cannot allow in our model", the authors point out, "is that superior courts are biased in the opposite direction to lower courts, because in this case higher error rates may be interpreted as reverse discrimination rather than evidence of mistakes by lower courts. We are not aware of a literature that documents such bias in opposite directions".
All death penalty sentences, in the US, are automatically appealed by higher state courts (the direct appeal) and, if they survive this stage, by state post-conviction courts and eventually they may be reviewed in federal habeas corpus petitions.
Collecting case-by-case data on the race of victims and defendants of all capital appeals from 1973 until 1995 (relying for the list on a dataset compiled by Fagan and Liebman in 2002, not containing racial information), Alesina and La Ferrara analyze 4,416 cases for the direct appeal and 531 for the habeas corpus and find evidence consistent with racial bias by lower courts. In the direct appeal sample, cases involving a minority defendant have an error rate of 37.7% if the victim is white and 34.7% if the victim is a minority, with a statistically significant difference of 3 percentage points; for habeas corpus cases the rates are respectively 37.5% and 28.4%, with a statistically significant difference of 9 percentage points. When defendants are white the differences are in the opposite direction, but are not statistically significant.
In the model Alesina and La Ferrara build, courts have to minimize a weighted average of two kinds of errors: sentencing to death an innocent defendant (Type 1 error) and acquitting a guilty one (Type 2 error) and the results of the test are consistent with a model where, when the victim is white and the defendant belongs to a racial minority, the weight attached to Type 1 error in courts' preoccupation is very low.
After conducting a number of robustness tests, the scholars state that "the result is not explained by differences in observable characteristics of the crime or of the trial, nor by the ideological orientation of appeal courts".