Systemic risk and financial contagion

• Applicant: Professor Dr.-Ing. Ulrich Meyer, since 12/2021
• Subject Area: Theoretical Computer Science
• Term: from 2020 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 411362735

The repercussions of the latest financial crisis took experts and academics alike by surprise. Especially the bankruptcy of the US investment bank Lehman Brothers has painfully revealed the dangers arising from the increasing complexities of interconnected finance operating at global scale and culminated in a near breakdown of the financial system. In this project, we want to develop novel and innovative methods for understanding and managing systemic risk arising from financial cross-holdings. To this end, we built on an established model of financial contagion and extend it along several interrelated directions. First, using a combination of theoretical analysis and computer simulations, we exploit that financial contracts can be valued as derivatives. In the case of single firms, the well known Merton model has introduced this idea and since been developed into an industry standard for credit risk management. In the context of financial networks, this connection has not been explored previously. Here, we intend to study financial valuations from this perspective in order to derive novel and theoretically well-founded measures of financial contagion and systemic risk insurance. Secondly, in collaboration with Deutsche Bundesbank, we plan to fit modern statistical models of networks to financial data. In particular, geometric models such as hyperbolic random graphs are well-known to reproduce degree distributions and clustering of many social networks, but have not been applied to financial networks. Furthermore, using Bayesian methods for latent variable models will allow us to generate artificial networks, with realistic structures matching real financial networks, in a privacy preserving fashion. This is especially important to cope with the strong confidentiality requirements of financial data which severely restricts access and free exchange of research data. Last, but not least, we discuss implications for effective management of systemic risk. In this context, we consider strategic aspects of financial contagion as well as algorithms for generating stress testing scenarios. Especially seemingly small shocks which nevertheless lead to system wide losses pose a major challenge in this context. Here, we propose to adapt techniques for patient zero problems in epidemics in order to reason back from the crisis scenario to shocks which could have triggered them. In the end, effective financial regulation requires a combination of precise measures of systemic risk and realistic models taking into account strategic considerations of individual financial actors.

• DFG Programme: Research Units
• Subproject of: FOR 2975: Algorithms, Dynamics, and Information Flow in Networks
• Ehemaliger Antragsteller: Professor Dr. Nils Bertschinger, until 12/2021