This paper offers a methodological approach for estimating classification error in police records then determining the statistical accuracy of official crime statistics reported to the Uniform Crime Reporting (UCR) program. Classification error refers to the mistakes in UCR statistics caused by the misclassification of criminal offenses, for example recording a crime as aggravated assault when it should have been simple assault. Statistical accuracy refers to the estimated true total of each crime type based on cancelling effect of undercounting and overcounting crime due to misclassifications. The population for the study consists of the 12 largest municipal police agencies in a mostly rural southeastern state. Based on a sample of 2,663 records, the authors illustrate the impact of classification error on the total population of reported offenses. Misclassifications result in overcounting and undercounting certain crimes. The true number of each crime type, as well as the aggregate Index Crime, Violent Crime, and Property Crime totals, is estimated based the evaluation of offsetting misclassifications. The findings show that certain UCR crime categories are greatly undercounted while others are overcounted. The index crime and violent crime totals are also significantly undercounted; however, when simple assault is added to the index and violent crime categories, the error in these aggregate numbers is reduced to less than 1 percent. The results provide a benchmark for assessing the statistical accuracy of the UCR data. (Published Abstract)