---
title: Expected shortfall estimation and capital coverage under overlapping scenarios
url: https://www.ml-quant.com/papers/arxiv/2610.41624/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-11-06
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: arXiv:2610.41624
source_url: https://arxiv.org/abs/2610.41624
featured: 2026-11-06
citations: unknown
topic: Risk, Credit & Banking
---


# Expected shortfall estimation and capital coverage under overlapping scenarios

When estimating expected shortfall from overlapping multi-day returns, the mismatch with standard independent-sampling theory causes systematic downward bias that weakens capital coverage, especially under heavy tails.

- Source: https://arxiv.org/abs/2610.41624
- Identifier: arXiv:2610.41624
- Released: 2026-11-02
- First featured: Quant Letter No. 138 (2026-11-06): https://www.ml-quant.com/issues/2026-11-06/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: Risk, Credit & Banking
- Authors: Martin Aichele, Igor Cialenco, Damian Jelito, Marcin Pitera

## Abstract (arXiv, CC0)

Expected shortfall (ES) is routinely estimated from a short sample of profit-and-loss (P&L) scenarios built from overlapping multi-day returns, while the non-parametric estimators applied to such data are designed under independent sampling. We study how this mismatch affects estimation accuracy and capital coverage, and how it interacts with serial dependence and heavy tails in the underlying P&L. Working within the coherent risk estimation framework, we compare six order-statistic-based ES estimators on fit, bias, and capital-coverage metrics. Following the Fundamental Review of the Trading Book (FRTB), we use a one-year estimation window and ES at the $2.5\%$ confidence level as a benchmark setup. We combine simulations with market data on twelve risk factors and ninety-nine equity portfolios. We show that overlapping sampling does not merely add noise but shifts ES estimates systematically downward, thereby weakening capital coverage. Serial dependence acts in the same direction, and heavy tails amplify both effects. More conservative weighting mitigates the underestimation but does not remove it. On market data, the confidence level at which the estimated capital covers realized losses roughly doubles in median under overlapping sampling for every estimator, while the $\textrm{ES}_{2.5\%}$-to-$\textrm{VaR}_{1\%}$ ratio, which links the FRTB base market risk capital metric to its pre-FRTB predecessor, changes comparatively little. Overlapping sampling can thus substantially understate ES-based capital. Since many banks apply a small set of standard estimators in their FRTB regulatory or economic capital model in a historical simulation approach, this understatement is likely to be common for many institutions, and combining simulation-based assessment of estimation risk with coverage-based diagnostics offers a practical method to detect it in model validation and monitoring.

## Related

- [A note on subadditivity of value at risks (VaRs): A new connection to comonotonicity](https://www.ml-quant.com/papers/doi/10-1017-jpr-2025-31/): The research reveals a new characteristic of value at risk (VaR), stating that its subadditivity holds for any confidence level only if the loss random variables are comonotonic.
- [Quantiles under ambiguity and risk sharing](https://www.ml-quant.com/papers/arxiv/2412.19546/): The study introduces Choquet Expected Shortfall, a new class of risk measures, and provides optimization algorithms and examples using financial data.
- [Time-Series Foundation AI Model for Value-at-Risk Forecasting](https://www.ml-quant.com/papers/arxiv/2410.11773/): The research highlights the superior performance of a time-series model, TimesFM, in forecasting Value-at-Risk, with fine-tuning further enhancing the results.
- [Optimal insurance design with Lambda-Value-at-Risk](https://www.ml-quant.com/papers/arxiv/2408.09799/): The paper studies optimal insurance solutions using the Lambda-Value-at-Risk model, revealing that a truncated stop-loss indemnity is ideal under certain conditions and discusses the effect of model uncertainty.
- [Adaptive Multilevel Stochastic Approximation of the Value-at-Risk](https://www.ml-quant.com/papers/arxiv/2408.06531/): The paper introduces a multilevel stochastic approximation algorithm that adaptively selects the number of inner samples to compute the value-at-risk of a financial loss, improving the previous scheme's complexity.
- [Risk sharing with lambda value-at-risk under heterogeneous beliefs](https://www.ml-quant.com/papers/arxiv/2408.03147/): The research investigates risk distribution among multiple parties using Lambda value at risk, offering formulas for optimal allocations under differing beliefs.