VaR, Expected Shortfall, and Tail Risk

VaR, expected shortfall, and tail-risk measures describe different parts of potential financial loss. Value at Risk estimates a loss cutoff, Expected Shortfall estimates average loss beyond a selected confidence level, and Tail Risk covers the broader exposure to extreme outcomes.

The same label can produce different numbers when horizon, confidence level, data, valuation, liquidity, or model assumptions change. Start with the decision and loss variable, then choose the measure that answers it. Conditional value at risk (CVaR) and conditional tail expectation (CTE) are covered on the Expected Shortfall page because they are commonly used as names for the same tail-average concept.

Key Takeaways

• VaR is a quantile, not a maximum loss.
• Expected shortfall adds information about modeled tail severity but remains an average, not a worst case.
• Cash-flow-at-risk and earnings-at-risk apply similar percentile or scenario ideas to business cash flow and earnings rather than only portfolio market value.
• Tail estimates need stress, liquidity, concentration, and model-risk analysis.
• Compare measures only when scope, horizon, confidence level, currency, valuation basis, and method are aligned.

Choose the Right Measure

For loss-bearing investment funds or internally allocated capital, see Risk Capital. The related phrase “capital at risk” is not a standardized percentile measure and needs a context-specific definition.

Example: Same Portfolio, Different Questions

Assume a one-day portfolio model reports 95% VaR of $2 million and 95% expected shortfall of $3.4 million.

• The VaR estimate identifies the modeled 95th-percentile loss cutoff.
• Expected shortfall estimates average loss in the modeled worst 5% of outcomes.
• Neither measure is the maximum possible loss.
• A stress test could still show a much larger loss under a specified price, volatility, credit, or liquidity shock.
• A liquidity review would ask whether the position can be hedged or sold within one day at modeled prices.

The figures become decision-useful only when connected to limits, capital, collateral, liquidity, and escalation.

What to Check

• Scope: positions, entities, products, risk factors, and risks included or excluded.
• Measurement: valuation date, horizon, confidence level, currency, and sign convention.
• Method: historical, parametric, Monte Carlo, scenario, or another approach.
• Data: lookback period, weighting, missing observations, stressed periods, and survivorship bias.
• Model: valuation, volatility, correlation, dependence, optionality, and aggregation assumptions.
• Liquidity: exit period, market impact, margin, collateral, and funding needs.
• Governance: model owner, validation, limit, breach process, and escalation authority.

Common Mistakes

• Treating VaR as the worst possible loss.
• Treating expected shortfall as a scenario maximum.
• Comparing numbers with different horizons or confidence levels.
• Assuming historical tail frequency will remain stable.
• Ignoring options, leverage, concentration, and forced selling.
• Reporting precise probabilities for events the model cannot represent reliably.
• Using a single metric instead of combining distribution-based measures with stress and liquidity analysis.

Educational Use

This section provides general financial education. It is not personalized investment, trading, banking, actuarial, regulatory, model-validation, capital, liquidity, or risk-management advice.

In this section

Choose a subsection first. Deeper term pages live inside each subsection, which keeps large topic hubs readable.