Survivability of networked agents with risk sharing

Abstract

Our society is governed by complex socio-technological networks, biological networks, and social networks. Among them, telecommunication networks, power grids, water distribution networks, transport networks or logistic networks are critical infrastructures that play a vital role in our modern society. It has become of paramount importance to understand and characterize the dynamic failures that might happen in these complex networks. We review, in detail, the previous work on cascade failure, and investigate the possibilities and constraints of creating decentralized and coordinated solution to the cascade failure problem. We demonstrate a few heuristics for mitigating cascading failure based on local-only knowledge actually do more harm than good. We then provide an effective method based on the concept of risk (load) sharing. We provide a simple micro-foundation based on coordinated incentives to absorb external shocks in order to survive collectively. The models of shock transfer are built up to investigate some stylized facts on how external or innate shocks tend to be allocated in the agent network, and how this allocation changes agents' failure probability. We analyze how risk sharing rules can actually affect the agent' resilience to external shocks. In particular, we focuses on: (i) how the allocation of risk is affected by the sharing rules, the network structure and the distribution of initial shocks among agents; (ii) how this allocation changes agents' failure probabilities.