A two-fold indicator of school performance and the cost of ignoring it

Traditional contextualised value-added models are useful tools for research and policy-making, but they overlook key assumptions necessary to fully understand academic performance. These models ignore significant sources of variation and fail to acknowledge the relationship between subjects. Using data from the Chilean Ministry of Education, this study fits a bivariate 5-level cross-classified model to assess the contribution of schools to pupils’ progress in Mathematics and Language. It is concluded that statistical models for school accountability need to fully account the variation between classrooms, primary schools and local authorities (in addition to variation between pupils and secondary schools), as well as the relationship between Mathematics and Language. This is to avoid undue lenience and/or harshness towards seemingly over-performing and under-performing schools.