Coherent Stress-Scenario Aggregation
Losses from two stress scenarios can overlap because the scenarios share the same underlying events. Adding their probabilities or losses directly can count the same paths twice. The Kwiatkowski–Rebonato framework uses linear programming to check whether the specified conditional probabilities fit any joint distribution. This can expose contradictory assumptions before they drive an aggregate risk estimate.
How it works
A set of stress scenarios can share underlying events. Their probabilities and losses cannot simply be added when events overlap. The Kwiatkowski–Rebonato technique uses linear programming to look for joint atomic probabilities consistent with specified conditional scenarios. Infeasibility flags contradictory inputs before they influence a portfolio-wide result.
Worked example
For a simple check use A with probability 20%, B with 15% and their intersection with 5%. The union has probability 30%, not 35%. If a proposed intersection were 18%, it would exceed the probability of B and no joint distribution could represent the inputs. More complicated conditional constraints can require a solver.
Validation and limitations
Document the basis of every probability and decide how to handle conflicting inputs. Inspect feasible ranges for aggregate loss when the conditions allow many distributions. A coherent probability set still needs relevant stress scenarios and credible loss estimates. Compare reverse stress testing.
Further reading
- Stress testing
Basel Committee · Supervisory guidance
- Macroeconomic Model Guide
Federal Reserve · 2025 supervisory documentation PDF
- Kwiatkowski and Rebonato, 2011 · DOI