From value-at-risk to expected shortfall: An agent-based analysis of market stability

Abstract

With the imminent entry into force of the Fundamental Review of the Trading Book (FRTB), banks around the world will adopt a common framework for measuring and managing market risk based on Expected Shortfall (ES). Previous research has shown that the reliance on homogenous models, particularly in the case of Value-at-Risk (VaR), can generate instability cycles arising from the synchronisation of investors’ actions. This paper investigates the potential of ES-based risk management systems to amplify market disturbances using an agent-based model populated by traders who manage their risk with an ES model and must reduce their positions whenever the ES-based risk of their portfolio exceeds a predefined limit. The results show that the widespread use of ES-based models can induce specific price dynamics that increase market instability. These dynamics, which we refer to as ‘ES cycles’, occur when a sufficient number of traders hit their ES limit simultaneously and are forced to reduce their positions at the same time. These collective selloffs depress prices, raise volatility, and force further liquidations, resulting in a self-feeding instability spiral. In addition, our results show that, although ES is expected to provide a better estimation of the tail risk, its adoption does not lead to any improvement with respect to the VaR case, as both models generate similar feedback mechanisms and instability patterns. However, the introduction of the stressed ES component as prescribed by the FRTB helps mitigate these instability episodes within our modelling environment.