This paper investigates the prevalence and predictors of teacher cheating in the Chicago Public Schools (CPS) using detailed administrative data. The authors develop an algorithm to detect cheating by analyzing unexpected test score fluctuations and suspicious patterns of answers within classrooms. Using data from 1993-2000, they estimate that serious cases of teacher or administrator cheating occur in at least 4-5 percent of elementary school classrooms annually. The study highlights that incentive systems, especially those with clear rules, can lead to behavioral distortions like cheating. Statistical analysis can help detect cheating despite attempts to keep it secret. The authors use two types of indicators to detect cheating: unexpected test score fluctuations and unusual patterns of answers within a classroom. Teacher cheating increases the likelihood of large, unexpected increases in test scores one year, followed by small gains or declines the next year. Cheating also leaves tell-tale signs such as blocks of identical answers, unusual correlations across student answers, or unusual response patterns on exams. Empirically, not every classroom with test score fluctuations and suspicious answer strings is cheating. The authors compare the observed distribution of test score fluctuations and suspicious answer strings to a counterfactual where no cheating occurs. They make three key assumptions: (1) cheating increases the likelihood of large test score fluctuations and suspicious answer strings, (2) if cheating classrooms had not cheated, their distribution of test score fluctuations and answer strings would be identical to non-cheating classrooms, and (3) the same pattern of correlation between test score fluctuations and suspicious answers observed for non-cheating classrooms in other parts of the distribution also holds in the upper tail. The authors find that cheating is more likely to occur in classrooms with large test score gains that are most likely attributable to cheating, and that students in such classrooms lose most of their gains the following year. Cheating is also correlated within classrooms over time and across classrooms in a particular school. The prevalence of cheating appears to respond to relatively minor changes in teacher incentives. The study concludes that teacher cheating is a significant issue that needs to be addressed.This paper investigates the prevalence and predictors of teacher cheating in the Chicago Public Schools (CPS) using detailed administrative data. The authors develop an algorithm to detect cheating by analyzing unexpected test score fluctuations and suspicious patterns of answers within classrooms. Using data from 1993-2000, they estimate that serious cases of teacher or administrator cheating occur in at least 4-5 percent of elementary school classrooms annually. The study highlights that incentive systems, especially those with clear rules, can lead to behavioral distortions like cheating. Statistical analysis can help detect cheating despite attempts to keep it secret. The authors use two types of indicators to detect cheating: unexpected test score fluctuations and unusual patterns of answers within a classroom. Teacher cheating increases the likelihood of large, unexpected increases in test scores one year, followed by small gains or declines the next year. Cheating also leaves tell-tale signs such as blocks of identical answers, unusual correlations across student answers, or unusual response patterns on exams. Empirically, not every classroom with test score fluctuations and suspicious answer strings is cheating. The authors compare the observed distribution of test score fluctuations and suspicious answer strings to a counterfactual where no cheating occurs. They make three key assumptions: (1) cheating increases the likelihood of large test score fluctuations and suspicious answer strings, (2) if cheating classrooms had not cheated, their distribution of test score fluctuations and answer strings would be identical to non-cheating classrooms, and (3) the same pattern of correlation between test score fluctuations and suspicious answers observed for non-cheating classrooms in other parts of the distribution also holds in the upper tail. The authors find that cheating is more likely to occur in classrooms with large test score gains that are most likely attributable to cheating, and that students in such classrooms lose most of their gains the following year. Cheating is also correlated within classrooms over time and across classrooms in a particular school. The prevalence of cheating appears to respond to relatively minor changes in teacher incentives. The study concludes that teacher cheating is a significant issue that needs to be addressed.