Coherent Stress Testing for Correlation Cascades: A Bayesian–Complex Systems Hybrid

Traditional stress tests often treat shocks as exogenous macro scenarios. Berre and Hoefman (2011) propose a Bayesian approach to coherent scenario generation, while Farmer et al. (2012) advocate complex-systems models that capture nonlinear feedback and network effects. Tabor and Zhang (2020) underscore that market correlation—rather than capital or liquidity levels—predicted run exposure around Lehman. This project integrates these strands into a supervisory tool that (i) generates priors over “correlation cascade” scenarios (e.g., sudden spikes in cross-bank betas and asset fire-sale elasticities), (ii) propagates them through an agent-based or mean-field model of banks and funds with endogenous deleveraging and common-factor shocks, and (iii) yields posterior distributions of losses and run probabilities. It challenges the norm by making correlation spikes a central, model-consistent shock, rather than a byproduct. Validation would involve retrodicting 2008 and the 2023 US bank failures (Ertürk 2023 on business-model pressures; Allen 2023 on VC/crypto linkages) to show improved explanatory power. The impact is practical: supervisors could “stress” the system’s propensity for synchronization and feedback, not just its point-in-time solvency.

References:

1. Interest Rates, Venture Capital, and Financial Stability. Hilary J. Allen (2023). Social Science Research Network.
2. Capital, Contagion, And Financial Crises. N. K. Tabor, Jeffery Y. Zhang (2020).
3. Coherent Stress Testing: A Bayesian Approach to Scenario Analysis and Stress Testing. Max Berre, Kevin Hoefman (2011).
4. Limits of bank regulation in financialised capitalism: lessons from the 2023 US banking crisis. Ismail Ertürk (2023). Law and Financial Markets Review.
5. UvA-DARE ( Digital Academic Repository ) A complex systems approach to constructing better models for managing financial markets and the economy. Doyne Farmer, M. Gallegati, C. Hommes, A. Kirman, P. Ormerod, S. Cincotti, A. Sánchez, D. Helbing (2012).