This thesis examines the validity of school value-added measures and the validity of arguments for their interpretation and use. The opening chapters review the development of school value added measures, existing evidence on their properties and validity and their current use in research, policy and practice. The empirical results are based on four studies using English National Pupil Database data and a large, nationally-representative dataset of teacher-assessed attainment data for English pupils aged from 7 to 13. The findings all relate to the properties of school value-added measures and the seriousness of a number of threats to their validity. The four empirical studies examine the following issues: observable bias and error, inter-method reliability when compared to estimates from a quasi-experimental regression discontinuity design, stability of school value added scores and of specific cohorts over time, and consistency of school value-added scores within cohorts and between different school cohorts at a single point in time. The closing chapters discuss the validity of value-added measures in general and in relation to the areas of use identified. Individually and collectively, the results advance understanding of numerous threats to validity and have substantial implications for the use of value-added measures in research, policy and practice.