The investigation takes an agent-based approach, which means that the key actors (banks, non-bank financial firms, and non-financial enterprises) are instantiated by computational objects executing strategies aimed to optimize objectives (profit) given available information. Modeling credit networks as multiagent systems affords heterogeneity: actors differing in objectives, information (access to data and observability of the environment), and capabilities (financial and technological resources). Evidence about systemic properties is derived by simulation, as models incorporating such heterogeneity and moreover accommodating complex information and fine-grained dynamics are analytically intractable. Since there are infinite combinations of agent behaviors one might consider—leading to different conclusions—it is necessary to adopt principled criteria for selecting salient agent strategies. This project employs empirical game-theoretic analysis methods for this purpose.
The first paper from this project, Strategic Payment Routing in Financial Credit Networks, was presented by Frank Cheng at the ACM Conference on Economics and Computation, Maastricht, in July 2016. The key contribution of that paper is the extension of credit networks to incorporate interest rates, producing a representation we call financial credit networks. Supporting interest rates requires several conceptual extensions, including maintaining a distinction between credit lines and debt holdings, and taking account of relative rates in routing payments. The paper presents an algorithm for determining the maximum payment flow between a pair of nodes that respects interest rate monotonicity along paths. We investigate the liquidity properties of financial credit networks, and explore tradeoffs between alternative payment mechanisms (i.e., rules for choosing payment paths) and liquidity. Finally, we consider the strategic implications of allowing agents to choose payment mechanisms on social welfare.
In ongoing work, we are applying financial credit networks for modeling systemic risk. Our focus is on replicating and extending prior models of the implications of alternative mechanisms for regulation of capital.
The project has supported two graduate student researchers to date (listed above).
- Accounting for strategic response in an agent-based model of financial regulation
- Strategic Payment Routing in Financial Credit Networks
- Strategic Modeling of Financial Credit Networks. US Treasury Office of Financial Research, 15 Sep 2016.
- Artificial Intelligence Meets Finance: Algorithmic Trading, Credit Networks, and Agent-Based Modeling. Dept of Informatics, University of Zurich, 24 Nov 2016.
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.