Expected shortfall estimation and capital coverage under overlapping scenarios

Expected shortfall is routinely estimated from short samples of overlapping multi-day returns, while the non-parametric estimators assume independent sampling. The authors compare six order-statistic estimators in a one-year window at the 2.5% level used by the Fundamental Review of the Trading Book, on simulations, twelve risk factors and 99 equity portfolios.

On market data, the confidence level at which estimated capital covers realised losses roughly doubles in median under overlapping sampling, for every estimator. More conservative weighting mitigates the bias but does not remove it, and the authors expect the understatement to be common among banks using standard estimators.