« L'économie des discriminations » peut-elle se passer d'une « philosophie économique des discriminations ? »

Since the 1950's, Nobel Prize economists such as Akerlof, Arrow, Becker, Heckman, Phelps and Stiglitz have analyzed discrimination in an economic framework, essentially in "positive" perspective. The two seminal models – Becker's model based on a "taste for discrimination" and Arrow and Phelps' theories of "statistical discrimination" – give two different answers to the implicit question of efficiency. Discrimination is inefficient – by assumption – in Becker's work because it's a « non-monetary cost », whereas discrimination could be efficient – in some cases – in statistical discrimination models. When discrimination is efficient, the normative basis disappears and no political recommendations could be made. The eficiency criterion is not sufficient to analyze discrimination both in a « positive » and in a « normative » perspective. The paper discusses two principles to renew this analysis: 1) individuals make choice using normative criteria; 2) the non-discrimination principle is lexically prior to an efficiency criterion