The Center for Research and Opinion Polls (CROP), as part of its activities, has welcomed a team of Technical Advisers of PASEC composed of Priscilla Gomes and Dr. Oswald Koussihouèdé, a research affiliate of CROP.
Dr Koussihouèdé has shared with a select group of CROP researchers his thoughts on evaluation methods, and has presented his own research on the effect of school size on pupils’ performance. He went over the good practices in experimental research and difficulties in their implementation in several cases. Those shortcomings do not imply that one should study only issues where those methods are perfectly applicable. He has stressed that such a approach would be detrimental to knowledge creation.
Dr Koussihouèdé’s research on school size contributes to informing policy makers on the optimal primary school size. The main question addressed by this research is the following: when should we build another school instead of allowing the extension of an existing one? To answer this question, Dr. Oswald Koussihouèdé uses three waves of data collected in Senegal in 2009, 2010, and 2011.
Since pupils cannot be randomly assigned to school of large or small size just for the benefit of this study, the research adopted non-experimental evaluation methods to build two statistically identical groups of pupils: those who attend small schools, and those who attend large schools. These two groups of pupils share the same observable characteristics, and it is thus possible to identify the effect of school size on pupils’ performance. The hypothesis (non-testable) of selection based on observables remains credible knowing a large number of control variables and, the econometric model (Value-Added Doubly Robust Estimations) bring a technical dimension to the research that allowed him to bypass possible violations of the identifying assumptions on the school-size effect.
The results of the research, that pass several sensitivity analyses, show that neither large schools are not favorable to pupils’ performance. Based on estimators’ variations in the sensitivity analyses, an optimal school size of 470-500 pupils is proposed. Even though these results are based on Senegalese data, we expect they are equally applicable to African educational systems that share the same characteristics and contextual background.