Architecture des réseaux interbancaires et gestion du risque de liquidité

In a standard stylised frame derived from Diamond Dybvig, banks operate within a network of debt contracts where liquidity shock distribution is unknown. Working in network enables banks to reduce the amount of liquid reserves and to decentralize a Pareto Optimal allocation while it is impossible if they stay isolated. However, this outcome depends on the architecture of the network, on the network participant number and on the cost structure. In a no cost framework, to decentralize first best outcome, networks have to exhibit one of two strong characteristics. It has either to have a « Small World property » which implies that banks must be bound together at very a short network distance, or, to have a strict regular topology. In a frame with positive cost, a single architecture both minimizes aggregate costs and decentralizes first best outcome. However, this topology, exhibiting unbalanced cost sharing among players, is not pairwise stable.