Measuring Self-Reported Wrongful Convictions Among Prisoners

Abstract

Estimate the frequency of self-reported factual innocence in non-capital cases within a state population of prisoners. We conducted a survey of a population sample of state prisoners who were asked to anonymously report their involvement in the crimes for which they were most recently convicted. To assess the validity of verifiable responses, prisoner self-report data were compared to aggregate conviction and demographic information derived from administrative records. To assess the validity of unverifiable responses, we developed a non-parametric test to estimate the probability of false innocence claims. We estimate that wrongful convictions occur in 6% of criminal convictions leading to imprisonment in an intake population of state prisoners. This estimate masks a considerable degree of conviction-specific variability ranging from a low of 2% in DUI convictions to a high of 40% in rape convictions. Implausible or false innocence claims are estimated to occur in 2% of cases. The present investigation demonstrates that survey methods can provide bounded estimates of factual innocence claims within a discrete and known population. The resulting estimates, the first to formally separate claims of legal and factual innocence and to incorporate a formal measure of response plausibility, suggest that prisoners themselves are very often willing to self-report the correctness of their convictions. At the same time, a considerable minority indicate that procedural weaknesses with the administration of justice occurred in their cases. And, a distinct minority, with considerable offense variation, maintain that they are completely innocent of the charges against them.