From agent-based modeling to actor-based reactive systems in the analysis of financial networks

Abstract

We present a new framework for the analysis of financial networks, called Actor-based Reactive Systems (ARS), that pushes further the Agent-Based approach (ABM) by resorting to ideas coming from the study of distributed systems in computer science. Two distinctive features, namely a fundamentally different management of time and a fully decentralized control logic, have a profound impact in terms of expressiveness of analysis, flexibility of modeling, and efficiency of experimentation. To illustrate the feasibility of the framework, we develop a realistic case study by analyzing the systemic risk of a model of the European banking network with a nontrivial contagion procedure, that combines an initial asset shock with the negative feedback loop triggered by asset fire sales. We show that, compared to ABMs, ARSs bring about finer-grained analyses, with a greater degree of heterogeneity and adaptivity of economic agents. Moreover, the very low computational cost and the detailed account of the system’s execution support the design and the development of very flexible stress tests to rapidly experiment with many hypothetical scenarios in a test-oriented style.