​Longer-Term Effects of Teachers: Evidence From a Random Assignment of Teachers to Students

Abstract

A critique of using teachers’ value-added (TVA) to measure their effectiveness is that teachers’ assignment to classes is not random, often involving endogenous sorting to match teachers and students. If valid, TVA measures and their estimated effect on students’ outcomes are biased. In this paper, we exploit an institutional setting of an education system in which high school teachers of all subjects are randomly assigned to classes of 11th and 12th graders. We use data from a representative sample of high schools in Greece to show that the same teacher’s TVA measures are highly persistent across time, different classes, and grades. Second, we estimate the effect of TVA on student outcomes during high school and higher education. We find that high-school students assigned to high-TVA teachers have higher test scores on high-stakes national exams, which are important for admission to higher education; are more likely to complete high school on time; are more likely to pursue postsecondary education; are more likely to enroll in universities than in technical or vocational schools; and win prestigious state scholarships for their undergraduate studies. The effect is also evident in admissions rates at higher-ranked universities, a more preferred study program, and expected wages. Remarkably, higher TVA in high school also affects the choice of field of study at university. The effect size of these impacts is meaningful economically. For example, a 1 SD improvement in TVA raises normalized average test scores by approximately 0.20 SD in 11th and 12th grades, and it increases the likelihood of continuing university education by 7 percentage points. These effects do not vary by the gender of students and are robust to various conditioning variables. We show that students assigned to higher-TVA teachers reduce their misconduct and truancy-related (unexcused) student class absences, which indicates higher motivation and engagement. We also show that high-TVA teachers, primarily male, are more likely to exit from schools and this is due to the mobility dynamics of teachers in low-performing schools. Joint paper with Victor Lavy (University of Warwick, United Kingdom)