Résumé en anglais

The concept of a barrier to entry has been discussed least since Bain [1956] with important contributions by Spence [1977], Dixit [1980] and Milgrom-Roberts [1982]. The more recent discussion is synthesized in the contributions to a dedicated session at the 2004 AEA meeting. Yet, a 'barrier to entry' remains a surprisingly elusive concept, which even accomplished theorists fail to define in an unequivocal manner. This article shows that past and current contributions to the subject fail to resolve the issue, because they insist on analyzing barriers to entry in the context of identical firms in homogenous goods industries. This inevitably leads to logical inconsistencies as each theorist provides his own ad hoc definition of a barrier to avoid Bertrand competition. We show that the notion of a barrier to entry has economic pertinence only if interpreted as a fully sunk cost in the form of a unique, indivisible, non tradable factor of production. This, however, implies monopolistic competition with welfare implications of a barrier to entry being positive as well as negative. Regulators thus need to ensure the ability of all competitors to create their own welfare-enhancing 'barriers to entry' in a dynamic context rather than to focus on their existence in purely negative and static terms.