Banking & Market Risk7 min read

VaR vs Expected Shortfall: Market Risk Measurement

By Jonas Adam Mohamed Osman, known as Yonas Osman

CEO of Quantica Risk

60 Rue François Ier, 75008 Paris, France

Published 15 October 2026

Why Basel shifted market risk capital from value-at-risk to expected shortfall, and what the change means for tail risk management in practice.

Executive answer

Value-at-risk estimates a loss threshold that will not be exceeded at a given confidence level over a set horizon, but it says nothing about how severe losses could be beyond that threshold — a blind spot the 2008 financial crisis exposed starkly when tail losses at several institutions far exceeded what 99% VaR had suggested was plausible. Expected shortfall, now the basis for FRTB market risk capital, instead averages losses in the tail beyond the VaR threshold, giving a more complete picture of potential severity and, crucially, satisfying the mathematical property of subadditivity that VaR lacks.

The practical consequence for banks is that expected shortfall is more sensitive to fat-tailed, extreme market moves and to correlation breakdowns during stress, meaning portfolios that looked well-diversified under VaR can show materially higher risk under expected shortfall if their tail dependencies were understated. This has required banks to overhaul risk models, data requirements, and in some cases trading strategies built around VaR-based risk limits.

The subadditivity problem that drove the shift

VaR's lack of subadditivity means that the VaR of a combined portfolio can, in specific circumstances involving non-normal return distributions, exceed the sum of the VaRs of its component parts — a mathematically awkward property for a risk measure meant to reward diversification. Expected shortfall does not have this flaw, making it a more theoretically sound basis for aggregating risk across trading desks and for setting coherent risk limits at the portfolio level, which was a key academic and regulatory argument behind the Basel Committee's decision to mandate it under FRTB.

Beyond the mathematical property, expected shortfall's focus on the average of tail outcomes rather than a single threshold also better aligns capital requirements with the actual loss banks would face in a genuine stress event, since by definition most realised losses during a crisis exceed the VaR threshold that was supposed to bound them 99% of the time — the entire point of a tail risk measure is how it behaves in exactly that remaining 1%.

Calibration to stressed periods and liquidity horizons

FRTB calibrates expected shortfall using a stressed observation period specific to each risk factor class, rather than a single historical lookback window, to ensure capital reflects what would happen under genuinely adverse market conditions rather than being understated during calm markets, as VaR-based Basel 2.5 capital was criticised for allowing in the run-up to 2008. This stressed calibration is more conservative but also more stable through the cycle, reducing the procyclicality where capital requirements fall during benign markets precisely when risk-taking is building up.

The framework also incorporates varying liquidity horizons by risk factor type, recognising that less liquid risk factors — certain credit spreads, emerging market currencies, structured product inputs — take longer to hedge or exit in a real stress scenario, so their expected shortfall contribution is scaled to a longer assumed holding period than a liquid government bond or major currency pair.

Practical implications for risk limits and model validation

Risk limit frameworks built historically around VaR thresholds need redesigning around expected shortfall, since the two measures do not move in lockstep — a portfolio with concentrated tail exposure can show moderate VaR but high expected shortfall, meaning VaR-based limits alone could have allowed that concentration to build undetected. Trading desks and risk committees need dashboards that present both measures together during the transition period, since historical risk appetite statements and board reporting calibrated to VaR language still need translation for comparability.

Model validation also becomes more demanding under expected shortfall, since backtesting a tail-average measure directly is statistically harder than backtesting a single quantile threshold; the Basel Committee addresses this by retaining VaR-based backtesting at the 97.5% and 99% levels as a validation tool even though expected shortfall is used for the actual capital calculation, creating a two-measure validation regime banks must manage in parallel.

Conclusion

The shift from VaR to expected shortfall reflects a hard-won lesson from 2008: a risk measure that ignores the severity of tail losses beyond its threshold cannot adequately capitalise genuine crisis scenarios. Expected shortfall is more complex to calculate, validate, and communicate, but it aligns capital far more closely with the losses banks actually experience when markets move violently.

Jonas Adam Mohamed Osman, known as Yonas Osman, CEO of Quantica Risk, reminds market risk teams that no single statistic replaces judgment about tail scenarios, and expected shortfall should complement rather than substitute for scenario-based stress testing of genuinely extreme, low-probability events.

Primary sources

value at riskexpected shortfallmarket risktail riskFRTB