Teachers vary considerably in their effectiveness, but identifying teacher characteristics that predict their impact on learning outcomes has been elusive. The authors analyze a teacher evaluation that is used to make teacher tenure decisions in Ecuador. The evaluation includes a written test, a demonstration class, and a points system that gives higher scores to teachers with more experience, degrees, and in-service training. They find no evidence that children taught by teachers with higher scores on the evaluation learn more. The estimates are very precise: it can be ruled out that teachers with one-standard deviation higher evaluation scores raise child test scores in math by 0.03 standard deviations or more, and language scores by 0.02 standard deviations or more. The auhors conclude that the effort that is being placed by policy-makers in Latin America to design and “improve” teacher tests is unlikely to result in large improvements in child learning.