Dr. Victor Cardenas-Santiago, University of British Columbia
Title: Integrating physical climate risk into existing credit models: a methodology for selecting and using catastrophe-model outputs
Date: Friday, November 27th, 2026
Time: 1:30PM (PDT)
Location: ASB 10900
Abstract: This paper develops a methodology for integrating physical climate risk, as estimated by catastrophe models, into existing credit-risk models and for determining the mathematical guarantees of the resulting credit metrics. Starting from the bank’s target metric, its information-sharing constraints and the outputs the provider can supply, the methodology specifies which catastrophe-model outputs to request, how to transform and aggregate them within the existing credit architecture, and whether the result meets the stated requirement or further information is needed, without access to the provider’s model internals. Each result is labelled as exact with respect to the provider’s model, exact under a stated hypothesis, an upper bound under any dependence, a certified approximation, or an uncertified estimate, using known comonotonic identities and an elementary bracket of the expected shortfall. An evaluation on a compound flood model of the Fraser River delta shows what each combination of information and hypotheses guarantees, how compensating errors hide inside a tolerance, and how the bracket directs the next information request.