Accuracy and Preferences for Legal Error

Legal procedures used to determine liability trade-off type-1 errors (e.g., false convictions) against type-2 errors (e.g., false acquittals). After noting that people's relative preferences for type-1 errors (compared to type-2 errors) appear to be negatively correlated with technological advancements, we study how the accuracy of evidence collection methods may affect the trade-off between these two errors. Counter-intuitively, we find that under some conditions greater accuracy may result in a higher probability of type-1 error (or type-2 error) maximizing deterrence. Then, assuming both errors are decreasing in accuracy, we characterize the type-1 error that emerges under electoral pressures (when the median voter's preferences are implemented): convictions occur more often than is socially optimal, but less often than is necessary to maximize deterrence. Moreover, as the harm from crime increases, the median voter becomes less tolerant of type-1 errors as the legal system's accuracy increases. We also show that, because the median voter is less averse towards type-1 errors than the average citizen, an increase in accuracy may reduce welfare.