Machine Learning for Stress Testing: Uncertainty Decomposition in Causal Panel Prediction
This framework provides a more robust and transparent method for regulatory stress testing, benefiting financial institutions and regulators by offering a clearer understanding of prediction reliability and uncertainty under stress scenarios. It is an incremental improvement to existing stress testing methodologies.
This paper addresses the challenge of projecting credit losses under hypothetical macroeconomic scenarios, a causal problem often treated as prediction. It introduces a framework that decomposes uncertainty into estimation and confounding components, providing a clear understanding of what can be learned from data versus what requires assumptions. The framework was validated through simulations and semi-synthetic experiments, including a Covid retrospective.
Regulatory stress testing requires projecting credit losses under hypothetical macroeconomic scenarios -- a fundamentally causal question typically treated as a prediction problem. We propose a framework for policy-path counterfactual inference in panels that transparently separates what can be learned from data from what requires assumptions about confounding. Our approach has four components: (i) observational identification of path-conditional means via iterated regression, enabling continuous macro-path contrasts without requiring a control group; (ii) causal set identification under bounded confounding, yielding sharp identified sets with interpretable breakdown values that communicate robustness in a single number; (iii) an oracle inequality showing that recursive rollout error is governed by a horizon-dependent amplification factor, providing a concrete answer to how far ahead one can reliably predict under stress; and (iv) importance-weighted conformal calibration bands with diagnostics that quantify extrapolation cost and trigger abstention when coverage guarantees degrade. The final output is a three-layer uncertainty decomposition that cleanly separates estimation uncertainty from confounding uncertainty. We validate all results through simulation and semi-synthetic experiments with real unemployment data, including a Covid retrospective demonstrating the framework's diagnostic value under extreme scenarios.