Institute for Mathematical Behavioral Sciences

The average accuracy of jury verdicts can be studied systematically and empirically even though the correct verdict may be unknown. The key? Obtain a second rating of the verdict (the judge's) as in the recent National Center for State Courts (NCSC) study of criminal cases from four jurisdictions in 2000-01. That study, like the earlier Kalven-Zeisel study, showed only modest agreement between the judge and jury. Estimates of jury accuracy are easily developed from the judge-jury agreement rate; under certain conditions the estimates of accuracy are optimistic, and it is suggested that those conditions hold. Numerical estimates of jury accuracy are presented. Of perhaps greater interest are estimates of false conviction rates and false acquittal rates, which are developed with the use of loglinear latent class models. Those models depend on stronger assumptions than the estimates based on agreement rates, and sources of uncertainty in the estimates are discussed. The estimates of the false conviction rates and false acquittal rates lead to questions about the appropriate balance of errors.