Stress-Testing with Fully Flexible Causal Inputs

Propagating causal stress-tests or contagion on selected risk factors to all the risk drivers is a challenging task. We use Entropy Pooling by Meucci (2008) to address this issue. Our causal stress-tests comprise, but are not restricted to, stress-testing Bayesian networks. We detail the theory and we present a case study: stress-testing a market driven by swap curve, credit spreads, stock market return, stock market volatility, currency strength, inflation, and commodities. Fully commented code supporting the empirical analysis is available for download