COMPASS

Studies aiming to understand school segregation have traditionally focused on decision making of individual households/parents (e.g., discrete-choice analysis, interviews) or macro-level patterns such as changes in levels and trends of segregation. However, despite the wealth of knowledge about school segregation and the many policy interventions that have been proposed to counteract it, segregation continues to plague educational systems across the world. Therefore, more recent studies have argued that school choice is a complex system of which the systemic behaviour can only be fully understood by viewing the system as a dynamic, interacting collective of individuals. Existing complexity approaches (e.g. Agent-Based Models, ABM) in this field are highly abstract theoretical models, which offer insight into fundamental dynamics, but are unable to predict real world dynamics . This is often due to the challenge of using real world data to parametrise the models. The challenge we aim to address is to calibrate an ABM on empirical, micro-level (CBS) data to better understand the macro-level phenomenon of school segregation. However, how to (efficiently) connect the model to the data and execute hundreds of thousands of simulations to facilitate empirical calibration (computationally costly) are still open questions for the ABM community. To our knowledge, if successful, this will be the first fully calibrated ABM in school choice or even in the social sciences. Hopefully this allows us to improve our understanding of school segregation, but also to pave the way for other (social) scientists/projects via a blog post / paper and/or code release.