Since 2007, the Ecuadorian government has required teacher candidates to pass cognitive and knowledge tests before they are allowed to participate in merit-based competitions for tenured positions. We evaluate this policy by linking administrative teacher information to data from an experimental study that randomly assigned nearly 13,000 children to their teachers. We find that test-screened tenured teachers had a significant effect of at least 10.5 percent of a SD on language learning outcomes. Although the recruitment tests screened candidates with higher cognitive skills, the classroom practice instrument used in the competitions appears to have helped identify the most effective teachers.