Constructing benchmarks for monitoring purposes: evidence from South Africa

In information-rich environments, schools are often presented with a myriad of data from which decisions need to be made. The use of the information on a classroom level may be facilitated if performance could be described in terms of levels of proficiency or benchmarks. The aim of this article is to explore benchmarks using data from a monitoring system on secondary school level. Seventeen secondary schools, purposively sampled for maximum variation, participated in this project. Pupils from a random sample of 2 Grade 8 classes per school completed the assessments (n = 1706). Using a dichotomous Rasch model, person item distribution maps were generated for mathematics, and different difficulty levels were constructed from the items that corresponded to the ability levels of pupils as well as the reporting protocol of the National Department of Education. Implications for benchmarking and standard setting are discussed based on the results provided in the article.