Stress-Testing with Fully Flexible Causal Inputs
working paper

    Published On 01/03/2011
    Author Name : 
    Attilio Meucci
    Published Date: 
    Tuesday, Dec 7, 2010
    Last Update: 
    Wednesday, Dec 8, 2010

    We propose a methodology to stress-test a set of risk drivers under minimal information. This methodology applies the entropy-based “fully flexible views” approach in Meucci (2008) and a novel consistency algorithm to extend the Bayesian network approach in Rebonato (2010). Starting from a plausible market distribution of the drivers, we are able to stress any conditional probabilities.These conditional probabilities comprise, but are not restricted to, causal Bayesian networks. Furthermore, in our approach we can also stress-test expectations, volatilities, correlations, quantiles, medians, etc.