A Methodology to Measure Impacts of Scenarios Through Expected Credit Losses

In this paper, we present a methodology for measuring the impact of scenarios on the expected losses of exposures by leveraging the existing provisioning infrastructure within financial institutions, where scenario effects are captured through changes in probabilities of default. We then describe how to design and implement a scenario test where risk drivers are given for standardized groupings of exposures, and the groupings are defined based on common features of the exposures. The methodology presented served as a theoretical foundation for the standardized climate scenario exercise conducted in 2024 by the Office of the Superintendent of Financial Institutions of Canada and Quebec’s Autorite des Marches Financiers.

💡 Research Summary

The paper proposes a practical methodology for quantifying the impact of macro‑level scenarios on the expected credit losses (ECL) of financial institution portfolios by leveraging the existing provisioning infrastructure mandated under IFRS 9 and CECL. The authors argue that because provisioning already requires the calculation of lifetime expected losses using probability of default (PD), loss‑given‑default (LGD) and exposure‑at‑default (EAD) forecasts, a scenario analysis can be performed simply by adjusting the PD (and optionally LGD) inputs according to scenario‑specific risk drivers.

The core technical contribution is the definition of a generic “scenario operator” that transforms baseline annual PDs into scenario‑adjusted PDs. The operator is implemented as a logarithmic add‑on to the conditional PDs, preserving the survival probabilities of the exposure while allowing a systematic increase (or decrease) in default risk driven by the scenario. The authors provide recursive formulas to decompose unconditional PDs into annual conditional PDs, apply the operator, and then reconstruct unconditional scenario‑adjusted PDs for use in the standard ECL formula:

ECL_sc = Σ_k w_k Σ_t PV