The Evolution of CCP Risk Management: a Tale of Automation, but Artificial Intelligence?

Abstract

The history of CCP risk management is one of gradual evolution, as a result of experience and experimentation. From early callable capital and basic forms of initial margin over 100 years ago to the sophisticated risk systems nowadays, it is hard to disassociate the evolution of CCP risk management from innovations in technology, developments in mathematical/statistical theories, and the necessity of agents to respond to new economic incentives and challenges. The recent resurgence of artificial intelligence (AI) represents an important step on this evolution. However, these developments now also pose the question of whether AI will disrupt the natural course of automation observed in CCP risk management or if, in fact, AI will simply add to it. The objective of the paper is to reflect on this type of question, aiming to understand whether the recent developments could be the preceding phases of another AI winter. Using an assessment framework that searches for AI-type characteristics in existing risk functions performed by the CCP, the main conclusion of the paper is that there are some areas like credit risk, custody risk, and investment risk that seem to have the features for efficient AI implementations. The supporting argumentation is that these risk functions rely typically on large amounts of mostly public information, to some extent not theoretically coherent across all its components, with models that combine quantitative and qualitative inputs. The paper's conclusions fall short of saying other CCP risk functions will not be impacted but, from a systematic point of view, these seem more prone to continue on their gradual automation trajectory.

Keywords: JEL Classification: E44, G01, G23, G28 Artificial Intelligence, Automation, Central Counterparties, Financial Regulation, Derivatives Markets, OTC derivatives, Risk Management, Systemic Risk