Compass-school
CO
This Python implementation tries to model school choice and resulting school segregation based on the work of Schelling (1971) and Stoica & Flache (2014).
Updated 5 months ago
5
Image from https://www.pexels.com/photo/children-sitting-on-brown-chairs-inside-the-classroom-4019754/
School segregation is potentially problematic, but hard to address. It results from the school choices of multiple interdependent individuals in a complex geographical and social environment.
Theoretical work showed that school segregation can be an unintended consequence of individual school choices. However, to know whether this applies to real empirical settings and which policies could be effective to mitigate school segregation, computationally highly intensive models are required.
Our Agent-Based Model (ABM) developed in the Computational Modelling of Primary School Segregation (COMPASS) project has many parameters. For exploring the full parameter space, conduct global sensitivity analysis, parameter sweeps, quantify uncertainty, policy simulations and empirical calibration, the model needs to run millions of times.
This eScience grant made it possible to run the model as efficient as possible for the full Amsterdam case, by reducing memory allocation and increasing the speed of calculations. We can now simulate primary school choices of roughly 70,000 households choosing schools based on distance and social composition. We can thus disentangle some of the complexity in mechanisms behind school segregation and explore several policy scenarios. Additionally this is a crucial step to incorporate more empirical realism in the ABM by connecting it to individual level data.