It has become commonplace to measure teacher quality in terms of teacher value-added. Operationally, this means evaluating teachers according to the learning gains of students on various achievement tests. Existing research consistently shows large variations in teacher effectiveness, much of which is within schools as opposed to between schools. The policy implications of these variations are dramatic. But the underlying statistical modeling has become the subject of intense research, in part because of this direct use of value-added measures in policy discussions.