Since 2007, the Ecuadorian government has required teacher candidates to pass national skill and content knowledge tests before they are allowed to participate in merit-based selection competitions for tenured positions at public schools in an attempt to raise teacher quality. We evaluate the impact of this policy using linked administrative teacher information to data from a unique experimental study where almost 15,000 kindergarten children were randomly assigned to their teachers in the 2012-2013 school year in Ecuador. We find positive and significant effects of test-screened tenured teachers of at least a 0.105 standard deviation for language and a 0.085 standard deviation for math, which persist even after controlling for teacher education, experience, cognitive ability, personality traits and classroom practices.