{
  "id": "16a617b6-63d0-53fa-bed7-7d96e8667235",
  "slug": "expected-shortfall",
  "term": "Expected Shortfall",
  "aliases": [],
  "category": "Risk Management",
  "category_slug": "risk-management",
  "difficulty": "advanced",
  "definition": "Expected Shortfall (ES), also called Conditional Value at Risk (CVaR) or Expected Tail Loss (ETL), is a risk measure that quantifies the expected loss of a portfolio given that the loss exceeds the Value at Risk (VaR) threshold at a specified confidence level, making it a coherent risk measure that captures the severity of tail losses rather than merely their probability. Unlike VaR, ES satisfies subadditivity—the ES of a combined portfolio is always less than or equal to the sum of individual ES values—making it theoretically superior for portfolio risk aggregation.",
  "key_takeaways": [
    "ES is the average of all losses worse than the VaR at a given confidence level—it answers 'how bad is bad?' beyond the VaR threshold.",
    "The Basel III/IV framework for market risk capital (FRTB) mandates use of a 97.5% ES to replace the previous 99% VaR standard, reflecting ES's superior tail risk capture.",
    "ES is sensitive to the tail distribution assumption: under fat-tailed distributions, ES can be substantially larger than under normal distribution assumptions.",
    "Computing ES by historical simulation involves averaging the worst 2.5% (or 1%) of observed scenarios, making it computationally straightforward but data-intensive.",
    "ES-based portfolio optimization—minimizing ES rather than variance—produces portfolios with better tail risk control but requires more sophisticated optimization algorithms."
  ],
  "detailed_explanation": "The conceptual case for Expected Shortfall over Value at Risk rests on a fundamental limitation of VaR: it provides no information about the distribution of losses beyond its threshold. If a portfolio has a 1-day 99% VaR of $10 million, all that is known is that there is a 1% probability of losing more than $10 million in a day—the loss could be $10.1 million (a near-miss) or $500 million (catastrophic) with equal statistical implication under VaR alone. ES addresses this by averaging the losses in the tail, providing a complete characterization of tail severity that VaR ignores.\n\nMathematically, ES at confidence level α is defined as: ES_α = E[L | L > VaR_α] = (1/(1-α)) × ∫_{α}^{1} VaR_u du. For a continuous distribution, this is the conditional expectation of the loss random variable given that it exceeds the α-quantile. For a discrete historical simulation with 1,000 scenarios, the 99% ES is the average of the 10 worst scenarios—straightforward to compute but sensitive to the accuracy and representativeness of the historical sample.\n\nThe mathematical properties that make ES superior to VaR for risk aggregation are formalized in the theory of coherent risk measures developed by Artzner, Delbaen, Eber, and Heath (1999). A coherent risk measure must satisfy four axioms: monotonicity, translation invariance, homogeneity, and subadditivity. VaR fails the subadditivity axiom: it is possible to construct two portfolios A and B such that VaR(A+B) > VaR(A) + VaR(B), implying that combining portfolios increases measured risk—a counterintuitive result that undermines diversification logic. ES always satisfies subadditivity, meaning diversification always reduces or maintains portfolio ES.\n\nThe transition from VaR to ES in banking regulation under the Fundamental Review of the Trading Book (FRTB) framework represents a major shift in how banks measure market risk capital. FRTB replaces the previous 10-day 99% VaR with a 10-day 97.5% ES, which under a normal distribution is roughly equivalent to the 99% VaR but captures more tail severity under fat-tailed distributions. FRTB also introduces liquidity horizons that scale the ES to account for the time required to liquidate different asset classes—a forward-looking improvement over the uniform 10-day horizon applied to all positions under Basel 2.5.\n\nFor hedge funds, ES is used in internal risk management, stress testing, and investor reporting. A fund reporting 5-day 95% ES of 3% to investors is conveying that in the worst 5% of 5-day periods, the expected loss is 3% of NAV—a more complete risk communication than VaR alone. ES-based risk limits—such as 'do not exceed 2% daily 99% ES on any single position'—are increasingly common in multi-strategy hedge fund risk frameworks because they constrain the entire tail distribution rather than just the entry point.",
  "example": "A hedge fund runs a Monte Carlo simulation of 10,000 daily return scenarios for its portfolio. The 99% VaR is $2.5 million—meaning 100 of the 10,000 scenarios result in losses exceeding $2.5 million. Those 100 worst scenarios have losses of: ranging from $2.5M to $8.0M, averaging $3.8 million. The 99% Expected Shortfall is therefore $3.8 million—52% worse than the VaR. This distinction matters: if the portfolio is exposed to a severe tail event (e.g., a short volatility position during a market crash), the 100 worst scenarios might average $12 million, giving an ES of $12 million even if the VaR is only $2.5 million. A pure VaR framework would miss this tail severity; ES captures it, motivating appropriate position size reduction or tail hedging.",
  "formula": "ES_α = E[L | L > VaR_α] = (1/(1-α)) × ∫_{α}^{1} q_u(L) du; Historical ES_α = (1/n_tail) × Σ L_i for all i where L_i > VaR_α",
  "formula_latex": null,
  "interactive_type": "calculator",
  "calculator_id": "expected-shortfall",
  "related_terms": [
    "aggregation",
    "conditional-value-at-risk",
    "diversification",
    "haircut",
    "hedge-fund",
    "hedge-ratio",
    "hedging",
    "idiosyncratic-risk",
    "liquidity",
    "market-risk",
    "maximum-drawdown",
    "monte-carlo-simulation",
    "normal-distribution",
    "risk-limits",
    "skewness"
  ],
  "backlinks": [
    "aggregation",
    "basel-iv",
    "central-limit-theorem",
    "conditional-value-at-risk",
    "correlation",
    "credit-risk",
    "downside-risk",
    "drawdown",
    "fat-tailed-distribution",
    "fat-tails",
    "historical-simulation-var",
    "incremental-var",
    "marginal-var",
    "market-risk",
    "monte-carlo-simulation",
    "monte-carlo-var",
    "portfolio-optimization",
    "short-hedge",
    "skewness",
    "stable-distribution",
    "standard-deviation",
    "stop-loss",
    "value-at-risk"
  ],
  "cross_references": [
    "aggregation",
    "conditional-value-at-risk",
    "diversification",
    "hedge-fund",
    "hedging",
    "liquidity",
    "market-risk",
    "monte-carlo-simulation",
    "normal-distribution",
    "risk-limits",
    "stress-testing",
    "value-at-risk",
    "volatility"
  ],
  "tags": [
    "level:advanced",
    "cat:risk-management"
  ],
  "asset_classes": [],
  "regulators": [],
  "see_also": [],
  "sources": [
    "artzner-1999"
  ],
  "wordcount": 821,
  "checksum": "a2a6151071d26237",
  "version": "2026.05.03",
  "license": "CC-BY-4.0",
  "updated_at": "2026-09-07T02:15:24+00:00",
  "_links": {
    "self": "https://hedgefund.wiki/api/v1/terms/expected-shortfall",
    "jsonld": "https://hedgefund.wiki/api/v1/terms/expected-shortfall?format=jsonld",
    "markdown": "https://hedgefund.wiki/api/v1/terms/expected-shortfall?format=md",
    "graph": "https://hedgefund.wiki/api/v1/graph/expected-shortfall",
    "category": "https://hedgefund.wiki/api/v1/categories/risk-management",
    "schema": "https://hedgefund.wiki/schema/term.schema.json",
    "html": "https://hedgefund.wiki/#/terms/expected-shortfall"
  }
}