A rigorous backtest for ES cannot exist, but a good approximation might do the job
Banque Pictet quant explains a new backtesting method for expected shortfall
In this paper, the author uses a unique data set, containing the revealed risk preferences of 9235 subjects, elicited with four different methods, to estimate latent risk preferences.
This paper aims to reflect the current state of the discussion on the validation of market risk forecasts by means of backtesting.
Fissler, Ziegel and Gneiting investigate the role of elicitability in backtesting problems and show how comparative backtests can be implemented for expected shortfall
Three easy-to-implement methods for back-testing expected shortfall