Abstract
In many countries, policy makers struggle with the development of value-added indicators of school performance for educational accountability purposes and in particular with the choice whether school context measured in the form of student composition variables should be included. This study investigates differences between 7 empirical studies that have looked at changes in value-added indicators through the inclusion of school composition variables. Differences were found in the strength of the association between Type A (student-level covariates) and Type B (student- and school-level covariates) school effect models in various regions, varying between 0.74 and 0.99 for secondary education. Important differences were found between the estimate of a school’s effectiveness by using a Type A or a Type B value-added model in secondary education in The Netherlands, Belgium, Chile, and Australia and in US and UK primary education. However, few differences were found in other cases.