Report from November 20 GRI Research workshop, “Extreme Risk and Model Risk: A “Learning By Failing” Approach
13 November 2015
While regulators have required that financial institutions assess model risk, there is no accepted approach for computing such risks. In this presentation, Professor Bertrand Maillet proposes a remedy for this lacuna via a general measure for estimating model risk which involves adjusting imperfect risk forecasts by outcomes from back-testing results.
Professor Maillet’s presentation was followed by industry responses from discussants Dr. Mark Staley, VP of Model Risk Development at TD Bank Group and Dr. Olga Strelchenko, Director and Financial Engineer at McGraw-Hill Financial.
Dr. Staley’s presentation responded to Professor Maillet first by noting banks’ practices in light of regulatory requirements such as the use of the Basel Traffic Light approach. He then highlighted the potential for employing the same methodology in related domains, including capital planning and enterprise stress testing as well as counterparty credit risk modeling. Dr. Strelchenko argued that even for historical measures such as Historical Value-at-Risk, the impact of model risk is pronounced as practitioners face the struggle of filling the gaps of missing data. Questions raised from the floor included the possibility of incorporating operational risk into such quantitative models, the framework application of Incremental Risk Charge (IRC), Prof. Maillet’s view on best practices for extreme risk modeling, and his practical advice to reduce the model risk inherent in estimates of extreme risk measures.