In spite of its wide acceptation amongst academics and practitioners as the favored option for peacebuilding in multi-ethnic societies, power-sharing has a poor record in practice. This paper argues that these practical inefficiencies are related to weaknesses in the theoretical underpinnings of power-sharing, namely an implicit primordialist conception of ethnicity, an a-contextual approach of institutional design, and an elite-bias. It proposes to re-approach power-sharing theory through the lenses of a constructivist theory of ethnicity and a neo-institutionalist approach of institutions, which pays attention to the interaction of institutions with their environment and to citizen-elite linkages. The theoretical discussion is complemented by a plausibility probe which focuses on Burundi. Three arguments are developed: 1-The politicization of the ethnicity thwarts the prospects of building sustainable peace and democracy. “Success” of power-sharing can thus be defined as the depoliticization of ethnicity. 2-The politicization of ethnicity results from a process in which formal and informal institutions as well as violence play key-roles. 3-The depoliticization of the ethnic cleavage is possible if (a) sufficient guarantees for the political representation and the security of the various ethnic categories are established, and (b) incentives are provided for the mobilization of non-ethnic identities by citizens and elites.
