Economic analyses of elementary education in the United States: three essays

Abstract

This dissertation investigates three substantial issues in education policy. First, it examines whether or not unobserved student heterogeneity introduces bias into value-added models of teacher effectiveness. After developing a statistical test, the results indicate that the omission of typically unmeasured information on noncognitive skills generates statistically significant bias and results in the misclassification of teachers according to their value-added quintile.Next, the dissertation explores the characteristics upon which students are assigned to teachers within schools. In contract to prior studies that analyzed data from limited geographic areas, this research employs a nationally-representative data set for the first time. While the analysis largely confirms earlier studies, I find that some of the matching documented in earlier studies, such as on prior test scores, might be an artifact of the chosen model, and that the sorting might instead take place on variables that are correlated with test scores. Finally, the dissertation critically examines prior research on class size reduction in elementary grades and generates estimates of future labor market benefits based on several later outcomes. Though estimates of labor market benefits based on ACT scores are higher than the costs of class size reduction, estimates generated using other outcomes almost exclusively are not.