P&L Attribution and Backtesting

Sanjay Sharma and John Beckwith

“The longer we keep looking back in the rearview mirror, it takes away from everything that’s moving forward” – Dan Quinn

After overcoming the challenges of IMA model approval, each RTD has to clear the monthly hurdle of the PLA test. Failing this cumbersome test would put the desk in SA purgatory with a capital shock. Beyond that, and even in the lush garden of IMA, the backtesting weeds can be unsightly for senior management, and sap capital in the form of a higher capital charge multiplier. The PLA test is necessary to avoid the pitfalls of model divergence across trading desks and risk control but, as currently defined, it verges on impossibility. It has several theoretical shortcomings and practical challenges that have generated significant and justifiable criticism from banks and other industry stakeholders. The prescribed backtesting construct requires backtesting of VaR, which is not a core risk measure in FRTB but is is the basis for backtesting ES. It would require market data for stress periods that the markets have not faced since the GFC. The breach threshold framework is supported by a statistical construct to approximate an optimal balance between type I and