부산대학교 도서관

This paper reports on findings based on analyses of a unique dataset collected by ACT that includes information on student achievement in a variety of subjects at the high school level, which allow us to examine the relationship between teacher effect estimates derived from VAM specifications employing different student learning assumptions. Specifically we explore the impact of assuming student learning to be unidimensional, prior knowledge ignorable, and achievement not influenced by concurrent teachers (spillover). We find evidence that both the estimated effect size of teacher quality changes and the estimates of individual teacher performance vary depending on learning assumptions. In particular, teacher effects estimated assuming unidimensionality and ignorable prior knowledge results in significantly smaller effect size estimates than those generated from a more traditional lagged score model, as well as changes in inferences about teachers. Similarly, teacher spillovers across subjects are found to have some impact on primary teacher estimates. Results imply that estimating high school teacher effects without explicitly considering the underlying assumptions results in biased estimates of performance; the substantive relevance is a normative question, but, our findings strongly suggest that VAM specification at the high school level warrants further research. (Contains 6 tables and 35 footnotes.)