Implementing Value-Added Models of School Assessment

This paper considers value-added models of school assessment and their implementation in Poland. Value-added estimates can be very helpful for schools and policy makers who need a reliable way to control teaching effectiveness, or for parents who need information about school quality in their area. However, their usefulness depends on several statistical issues and specific decisions made during implementation. The paper discusses several value-added models and describes details of the solution implemented in Poland. Statistical problems are discussed according to their policy relevance. It is shown that what bothers statisticians is less important in practice than several problems encountered when one wants to apply these models to a policy relevant context. Problems of proper regression specification, omitted variables bias, and measurement error are discussed, but the ways value-added estimates could be published and used as policy evaluation tools are also presented. All this problems are discussed from a practical point of view using three years of experience in implementation of these methods in Poland.