ABSTRACT

The P&L Attribution (PLA) test was first proposed by the Basel Committee in October 2013. The test was introduced as a new component to the Internal Model Approach (IMA) for calculating Pillar 1 market risk capital. Over the subsequent years, the substance and the details of the PLA test were unfortunately never challenged, either by industry or academia, and the test duly became enshrined in the “final” Market risk standard released in January 2016. The PLA test is applied at desk level and hinges upon a monthly, pass-fail test involving the calculation of two ratios which are based on the mean and variance of the “hypothetical” (front-office) P&L and the “risk-theoretical” P&L. A desk which fails a fourth month in any rolling 12-month window automatically loses accreditation to use IMA and instead must revert to using the Standardised Approach (SA). ANZ’s work, with a co-author from academia, set out to analyse the PLA test using theoretical but not unrealistic mathematical assumptions about the statistical distributions of the hypothetical and risk-theoretical P&L. Our conclusion was that the PLA test was unlikely to be feasible in practice. Based on the assumed distributions for the daily P&L, the failure rates were draconian, and depending on how quickly a desk could remediate their input data and regain IMA accreditation, the likely outcome would be only a modest proportion of desks having IMA accreditation at any given time. In short, the PLA test would mean that, in terms of capital savings, the cost-benefit from a bank’s pursuit of IMA accreditation would be questionable.

ABOUT THE SPEAKER

Peter Thompson has worked in financial risk management for many years, with roles across market risk, as well as wholesale and retail credit risk. He has a strong interest in post-GFC reforms to the global prudential regulatory framework, in particular Pillar 1 risk analytics and the ongoing attempts to salvage the concept of RWA.

Sponsored by Commonwealth Bank

You can add yourself to the SFMW mailing list.