On aggregate model risk management: Focus on stress testing

Abstract

Nowadays, financial institutions, especially banks around the globe, have adopted and significantly relied on models of different kinds in their day-to-day operations in loan underwriting, hedging, trading accounts, business planning, stress testing, budgeting, performance reporting, risk management and new product exploration, and so on. Therefore, bad design or implementation of models can expose financial institutions and the financial system to considerable financial risk. Unlike common risk types such as credit risk, market risk and operational risk, however, which have been studied by academia and the financial industry for decades, effective management of model risk remains a new and hot-button issue. More specifically, many studies have been conducted on the identification of model risk via model validation. However there is still not an established industry standard practice for the measurement and mitigation of model risk. This paper proposes a practical framework to measure model risk, based on extensive research and industry practice experiences. The framework will potentially address the difficulties surrounding the quantification of financial model risk and aggregate model risk, while commonalities and interdependencies among models should be noted and measured.