This study estimates error rates in identification of upper elementary school teachers as low or high performing based on student test score gain data. The study develops error rate formulas for commonly-used performance measurement schemes that are based on OLS and Empirical Bayes estimators and value-added models, where educator performance is compared to the district average using hypothesis testing. Simulation results suggest that performance estimates are likely to be noisy using the amount of data that are typically used in practice—1 to 3 years. Type I and II error rates are likely to be about 25 percent based on three years of data and 35 percent based on one year of data. Corresponding error rates for overall false positive and negative errors for all teachers who are subject to misclassification are 10 and 20 percent, respectively. Lower error rates can be achieved by increasing the number of student achievement gain measures that are available for any teacher. School-level results also have less error.