Education researchers have long made inferences about grade retention from the grade distribution of same-aged students. Recent economics studies have followed suit. This paper examines the validity of the "below grade" proxy for retention using data from supplemental questionnaires administered in the U.S. Current Population Survey during the 1990s. I estimate that 21% of non-repeaters are below grade, while 12% of repeaters are not. Misclassification attenuates regression coefficients by 35% when the proxy is an outcome and by 65% when it is a regressor. The latter figure is a benchmark, as classification and regression errors are arguably correlated. Biases are likely substantial in other surveys and time periods.