School Governance and Learner Performance in Sub-Saharan Africa: A Neural Networks Approach

The aim of this paper is to analyse the impact of school governance on learner performance in Sub-Saharan Africa, in the face of persistent low performance in the region, revealed by the PASEC 2019 report. The study uses an econometric model followed by machine learning models (Regression Logistic, Random Forest, Extra Tress Classifier, Extreme Gradient Boosting, Artificial Neural Networks) to explore the relationships between school results and governance factors measured by school management, pedagogical practices and relations with stakeholders. The results show that artificial neural network models perform better than conventional approaches in terms of accuracy and explainability. Explainability by Shapley values shows that the quality of administrative and pedagogical management, benevolent school-student relations, and activities to promote the best students significantly improve performance. The study suggests capacity building for managers in order to improve the quality of administrative and pedagogical management. It also highlights the need to promote rigorous administrative governance, based on effective practices and adapted to local realities. In addition, specific strategies should be put in place to reward high-performing students, while encouraging professional collaboration between education stakeholders. Finally, a review of parental involvement practices is recommended in order to avoid inappropriate expectations likely to be detrimental to learners' performance.