A Statistical Study of the Revised FRTB's P&L Attribution Tests

Abstract

One of the key challenges implied by the upcoming regulatory framework for minimum market risk capital requirements, known as FRTB (Fundamental Review of Trading Book), is the PL attribution (PLA) tests. PLA tests constitute a game-changer in the way risk engine and models are assessed for effectiveness and accuracy by regulators to base the qualification for the ‘internal model approach’ for market risk capital. The initial design of these tests, however, has raised concerns from the industry that pointed to a punitive behavior and room for improvements. In this context, the Basel Committee has proposed a completely revamped test design in their March 2018’s Consultative Document. In this paper, we first provide a formal statistical analysis of the performances and the behavior of this newly proposed PLA test as compared to the old one. Our findings suggest that although the revised test design addresses many of the raised concerns, there are still some improvement opportunities, particularly in what relates to the conceptual treatment of hedged portfolios. We also provide a detailed statistical study to orient the selection of the homogeneous PLs distributions test to be retained among the offered alternatives.