The Ridge Backtest for Expected Shortfall : Properties and applications, for FRTB IMA and more

We illustrate the "ridge backtest" for ES (proposed in Acerbi and Szekely (2017, 2019)) in the context of the FRTB IMA, ES-based regulation which is coming into effect soon. This backtest is shown to have unique optimal properties that solve the longstanding puzzle on whether ES can be backtested at all. Besides providing a viable effective backtest, the proposed method provides richer information than traditional VaR backtests, in the form of prediction discrepancy metrics in monetary terms. This permits actionable capital management diagnostics and remediation, and opens the door to the advent of adaptive risk models. We discuss in particular the relevance that this method could have for model validation of FRTB IMAs.

Paper