Predictions of Future Dangerousness and Rates of Error

Press sources including the Houston Chronicle are reporting on a study released yesterday by the Texas Defender Service that casts doubt on expert predictions of future dangerousness in capital cases. The study followed 155 inmates labeled dangerous by prosecution experts during the sentencing phase of their trials. Only 8 of them, or about 5%, were subsequently involved in serious assaults.

Critics of the study point out that for at least part of the study period, 40 of the inmates were on death row in near-total lockdown, which is supposed to prevent assaultive behavior. There is also a difference, of course, between predicting dangerousness within a prison setting over a finite time period and predicting dangerousness in the population at large over a lifetime. The press reports don't say whether the study investigated what rate of assaultive behavior would be expected from a comparable prison population comprising inmates for whom predictions of expert dangerousness had not been made. We have no first hand knowledge of prison life. But mightn't it be expected that the number of assaultive inmates in general would, if anything, exceed 5%?