Interpreting Tests of School VAM Validity

Peer-reviewed Publication

Joshua Angrist, Peter Hull, Parag Pathak, Christopher Walters

May 2016

Public school districts increasingly look to the estimates generated by value-added models (VAMs) to measure school and teacher quality. A typical VAM compares average student test scores across teachers or schools while using regression models to control for students’ past scores and demographic characteristics. The resulting estimates serve as inputs into teacher retention and promotion policies, school report card systems, and decisions about which schools to close, restructure, or expand. This paper lays out the econometric theory behind lottery-based VAM validity tests.