Addressing premarket discrimination through geographically-targeted affirmative action: the “Conventions Education Prioritaire ” of Sciences Po

Premarket discrimination, by which individuals underperform in skills acquisition because they expect lower returns to skills, is thought to be an important driver of differences still prevalent between groups in the labor market. This paper analyses the effects of an affirmative action admissions program rewarding effort among disadvan-taged high school students. We use a unique experiment run since 2001 by Sciences Po (Paris), a French selective higher education establishment which signed special agree-ments with high schools from disadvantaged neighborhoods in order to admit their best students through a special procedure. We propose a model of education investment decisions with costly effort and analyze changes in individual effort given the distri-bution of talent when the probability and size of the reward to effort change.Using a difference-in-differences methodology, we quantify the effects on academic results in participating high schools and on their attractiveness. We find that the program had, overall, only small effects on high school-leaving examination results and that the high schools that benefited most from the program were those that had the highest propor-tion of deprived students within eligible schools. The attractiveness of these schools to students from higher socio-economic status families also increased. These findings imply that such geographically-targeted “affirmative action ” programs can potentially improve school results in deprived areas without addressing premarket discrimination for the most underprivileged youth, through increased social diversity and peer-effects. ∗Diagne: Department of Economics, UC Berkeley; Wasmer: Sciences Po Paris, OFCE, Crest, IZA and CEPR. We are grateful to the Direction de l ’ Evaluation, de la Prospective et de la Performance (DEPP) at the French ministry of national education for providing the data, as well as Richard Descoings, Cyril Delhay and XXX at SciencesPo for providing support and data access. We thank XXX for helpful comments and remarks. Elizaveta Archanskaia provided outstanding research assistance. 1