# Parametric VaR & Expected Shortfall Guide (2026) | QuantCalc

> Master parametric VaR formulas, Basel FRTB 97.5% Expected Shortfall, Cornish-Fisher skew adjustments, GARCH decay, and backtesting.

Canonical: https://quantcalctools.com/guides/parametric-var-expected-shortfall-fat-tail-calibration-guide
Last updated: 2026-11-14T19:30:01.339Z

**Publication:** QuantCalc Tools
**Category:** Parametric Var Expected Shortfall
**Published:** 2026-11-14T19:24:25.915Z
**Author:** Elena Rostova, FRM, CQF (FRM, Certificate in Quantitative Finance (CQF))
**Reviewer:** Not recorded

## Answer-first summary

Under Basel III MAR33, 97.5% Expected Shortfall replaces 99% Gaussian VaR to penalize tail kurtosis; this quantitative guide reconciles Cornish-Fisher quantiles, Student-t scaling, EWMA decay, and backtesting.

## Key takeaways

- Under pure Gaussian assumptions, 97.5% Expected Shortfall (2.3378 sigma) nearly matches 99.0% VaR (2.3263 sigma), but Student-t (nu=5) tails widen 97.5% ES by +23.64% to 2.8904 sigma.
- Cornish-Fisher 4th-order quantile expansions capture negative skewness and excess kurtosis in microseconds, inflating 99% VaR from 2.3263 sigma to 3.3814 sigma at S=-0.80 and Ke=4.50.
- Basel III MAR33 replaces flat 10-day square-root-of-time scaling with five liquidity horizon buckets (10, 20, 40, 60, and 120 days) to reflect asset-class exit illiquidity.
- Regulatory model validation requires joint Kupiec POF unconditional frequency tests, Christoffersen Markov independence tests, and Acerbi-Szekely Z2 tail-magnitude backtests.

## Related financial entities

- **Basel III FRTB 97.5% Expected Shortfall (ES) Tail Metric** (FRTB-ES-975) — license: Basel III FRTB 97.5% Expected Shortfall (ES) Tail Metric (FRTB-ES-975) refers to an institutional financial benchmark and regulatory standard monitored by BIS Basel Committee MAR33 (INTL), currently calibrated at 97.5% (Confidence Interval) as of October 2026.
- **Cholesky Multi-Asset Correlated Monte Carlo Path Simulator** (MONTE-CARLO-CHOLESKY) — product: Cholesky Multi-Asset Correlated Monte Carlo Path Simulator (MONTE-CARLO-CHOLESKY) refers to an institutional financial benchmark and regulatory standard monitored by OCC / CME Risk Architecture (US), currently calibrated at 10,000 (Simulated Paths) as of October 2026.
- **CME SPAN 2 Historical & Stress Value-at-Risk (VaR) Model** (CME-SPAN2-VAR) — product: CME SPAN 2 Historical & Stress Value-at-Risk (VaR) Model (CME-SPAN2-VAR) refers to an institutional financial benchmark and regulatory standard monitored by CME Clearing (US), currently calibrated at 6.26% (Initial Margin / Notional) as of October 2026.
- **Cornish-Fisher Skew & Kurtosis Modified Parametric VaR** (CORNISH-FISHER-VAR) — product: Cornish-Fisher Skew & Kurtosis Modified Parametric VaR (CORNISH-FISHER-VAR) refers to an institutional financial benchmark and regulatory standard monitored by GARP FRM Quantitative Standard (INTL), currently calibrated at 99.0% (Non-Normal Quantile) as of October 2026.

## Analysis

At a 99.0% one-day confidence threshold, standard Gaussian Value-at-Risk ($z_{0.99} = 2.3263$) understates empirical tail losses by 28% to 54% whenever multi-asset return distributions exhibit negative skewness ($S < -0.50$) and excess kurtosis ($K_e > 3.0$). Institutional risk engines governed by BIS Basel Committee MAR33 and UK Prudential Regulation Authority ([PRA](/regulator-index/prudential-regulation-authority-uk-173)) internal models rules therefore transition trading-book capital from 99% VaR to [Basel III FRTB 97.5% Expected Shortfall](/entities/frtb-es-975-basel-iii-frtb-97-5-expected-shortfall-es-tail-metric), where normal tail averages coincide ($2.3378\sigma$ vs $2.3263\sigma$) under Gaussian assumptions yet diverge sharply under fat-tailed regimes. You can benchmark baseline normal tail quantiles directly inside our interactive [Value at Risk (VaR) calculator](/calculators/value-at-risk-calculator) before layering non-normal moment expansions and liquidity-horizon multipliers.

## Gaussian Parametric VaR vs Coherent Expected Shortfall Axioms

Variance-covariance VaR defines a quantile boundary—the maximum loss not exceeded with probability $\alpha$ over holding period $T$—but tells a risk committee nothing about loss severity once the $1 - \alpha$ boundary breaks. Worse, Artzner, Delbaen, Eber, and Heath proved that VaR violates subadditivity when credit or short-option payoffs are skewed: merging two concentrated desks can mathematically produce $VaR(A + B) > VaR(A) + VaR(B)$, penalizing diversification. Expected Shortfall (ES), also termed Conditional Value-at-Risk (CVaR), integrates losses across the entire tail beyond $VaR_\alpha$:

```formula
Gaussian Parametric VaR & Expected Shortfall Baseline ($100,000,000 Portfolio)
Inputs:
 Position Notional (W) ............ $100,000,000
 Annualized Volatility (sigma_a) .. 18.40% (252 trading days)
 1-Day Horizon Volatility ......... 0.1840 / sqrt(252) = 1.15909% ($1,159,091)

1) 99.0% Gaussian 1-Day VaR (z_0.99 = 2.32635):
 VaR_99 = $100,000,000 * 2.32635 * 0.0115909 = $2,696,451

2) 97.5% Gaussian 1-Day Expected Shortfall (z_0.975 = 1.95996, phi(z) = 0.05844):
 ES_97.5 = W * sigma_1d * [phi(z_0.975) / (1 - 0.975)]
 ES_97.5 = $100,000,000 * 0.0115909 * (0.0584409 / 0.025) = $2,709,590

3) Student-t (nu = 5) 97.5% Standardized Expected Shortfall Comparison:
 Standardized t_5(0.975) = 2.57058 * sqrt((5 - 2)/5) = 1.99116
 ES_97.5(t_5, unit variance) = 2.8904 * sigma_1d = $3,350,237 (+23.64% tail surcharge)
```

Notice the calibration symmetry engineered by Basel MAR33: under pure normality, 97.5% ES ($2.3378\sigma$) sits within 0.49% of 99.0% VaR ($2.3263\sigma$). The moment return tails thicken to Student-$t$ with $\nu = 5$ degrees of freedom, unit-variance 97.5% ES jumps by 23.64% to $2.8904\sigma$, forcing trading desks to hold capital commensurate with extreme tail realizations. Our dedicated spoke on [Student-t Expected Shortfall degrees-of-freedom scaling](/guides/student-t-expected-shortfall-degrees-of-freedom-tail-scaling) derives the exact gamma-function density factor and variance normalization required when $\nu \le 6$.

![Institutional tail-risk calibration pipeline from raw return filtration to Basel FRTB capital validation](/api/figure?type=flow&title=Institutional%20Parametric%20VaR%20%26%20FRTB%20ES%20Calibration%20Pipeline&subtitle=Four-stage%20workflow%20linking%20conditional%20volatility%20filtration%2C%20non-normal%20quantile%20expansion%2C%20and%20regulatory%20backtesting.&steps=1.%20Volatility%20Filtration%3A%3AFit%20EWMA%20lambda%200.94%20or%20GARCH%281%2C1%29%20to%20strip%20clustering%7C2.%20Tail%20Quantile%20Model%3A%3AApply%20Cornish-Fisher%20moments%2C%20Student-t%2C%20or%20GPD%20EVT%7C3.%20Horizon%20%26%20Covariance%3A%3AScale%20MAR33%2010d-120d%20buckets%20and%20Cholesky%20PSD%20grids%7C4.%20Backtest%20Governance%3A%3ARun%20Kupiec%20POF%2C%20Christoffersen%20Markov%2C%20and%20Z2%20ES%20tests&metrics=GAUSSIAN%2099%25%20Z-SCORE%3A%3A2.3263x%3A%3ABaseline%20normal%20quantile%7CFRTB%2097.5%25%20ES%20FACTOR%3A%3A2.3378x%3A%3ANormal%20phi%28z%29%2F%281-alpha%29%20ratio%7CSTUDENT-T%20%28NU%3D5%29%20ES%3A%3A2.8904x%3A%3A%2B23.6%25%20fat-tail%20capital%20uplift%7CKUPIEC%20RED%20ZONE%3A%3A%3E%3D10%20hits%3A%3A250-day%2099%25%20VaR%20breach%20cap)

## Non-Normal Moment Expansions, Heavy Tails, and Extreme Value Theory

When risk managers need closed-form speed, [Cornish-Fisher Modified Parametric VaR](/entities/cornish-fisher-var-cornish-fisher-skew-kurtosis-modified-parametric-var) adjusts $z_\alpha$ via third and fourth cumulants. In [Cornish-Fisher quantile expansion math](/guides/cornish-fisher-var-skewness-excess-kurtosis-quantile-expansion), skewness $S = -0.80$ and excess kurtosis $K_e = 4.50$ lift the 99% multiplier from $2.3263$ to $3.3814$ (+45.35%) within Maillard monotonicity bounds.

For non-linear derivatives or extreme tails, three architectures lead 2026 risk desks:
- **Taylor Delta-Gamma**: [Delta-Gamma quadratic Taylor VaR](/guides/delta-gamma-var-quadratic-taylor-options-convexity-adjustment) captures negative option convexity via $\frac{1}{2}\text{tr}(\Gamma\Sigma)$.
- **Extreme Value Theory**: [EVT Peaks-Over-Threshold GPD tail ES](/guides/extreme-value-theory-peaks-over-threshold-gpd-tail-es-model) fits Pareto tails above threshold $u$.
- **Lognormal Horizon Bounds**: [Lognormal vs arithmetic VaR](/guides/lognormal-vs-arithmetic-return-var-large-drawdown-horizon-error) prevents impossible >100% long-only losses via $1 - \exp(\mu_h - z_\alpha\sigma_h)$.

## Conditional Volatility Dynamics and Multi-Day Horizon Scaling

Equal-weighted 250-day windows suffer abrupt Day-251 ghost drop-offs. Comparing [EWMA lambda 0.94 against GARCH(1,1) conditional volatility decay](/guides/ewma-lambda-094-vs-garch-11-conditional-volatility-var-decay) anchors multi-day forecasts to long-run variance $\sigma_{LR}^2$, while [Filtered Historical Simulation (FHS)](/guides/filtered-historical-simulation-garch-standardized-residual-var) resamples GARCH-standardized residuals and [Garman-Klass and Yang-Zhang OHLC estimators](/guides/intraday-garman-klass-yang-zhang-range-volatility-var-inputs) capture overnight jumps.

Scaling 1-day volatility via $\sqrt{T}$ fails under AR(1) persistence—an error derived in [square-root-of-time VaR serial correlation bias](/guides/square-root-of-time-var-scaling-serial-autocorrelation-bias) alongside our [Sharpe & Sortino ratio calculator](/calculators/sharpe-sortino-calculator). Under MAR33.4, [Basel FRTB 97.5% ES liquidity horizon bucketing](/guides/basel-frtb-975-expected-shortfall-liquidity-horizon-bucketing) replaces $\sqrt{10}$ with a cascade across 10, 20, 40, 60, and 120 days.

| Tail-Risk Model Specification | Primary Governing Benchmark / Entity | 1-Day Tail Multiplier (Unit $\sigma$) | Skew & Kurtosis Sensitivity | Primary Institutional Use Case |
| :--- | :--- | :--- | :--- | :--- |
| Gaussian Parametric VaR (99.0%) | [VaR Calculator](/calculators/value-at-risk-calculator) | $2.3263\sigma$ | Zero ($S=0, K_e=0$) | Liquid G10 FX & intraday desk limits |
| Gaussian Expected Shortfall (97.5%) | [FRTB-ES-975](/entities/frtb-es-975-basel-iii-frtb-97-5-expected-shortfall-es-tail-metric) | $2.3378\sigma$ | Coherent tail integral | Baseline MAR33 linear rates benchmarking |
| Cornish-Fisher VaR ($S=-0.6, K_e=3.5$) | [CORNISH-FISHER-VAR](/entities/cornish-fisher-var-cornish-fisher-skew-kurtosis-modified-parametric-var) | $3.0842\sigma$ | 4th-order cumulant polynomial | Fast multi-asset skew/fat-tail desk VaR |
| Student-t Expected Shortfall ($\nu=5$, 97.5%) | [FRTB-ES-975-STD](/entities/frtb-es-975-2026-benchmark-standard) | $2.8904\sigma$ | Symmetric power-law tail decay | Sovereign credit & EM FX capital sizing |
| CME SPAN 2 Historical + Stress VaR | [CME-SPAN2-VAR](/entities/cme-span2-var-cme-span-2-historical-stress-value-at-risk-var-model) | $6.26\%$ IM / Notional | Empirical crisis + HVaR blend | Cleared futures & options initial margin |
| Cholesky Correlated Monte Carlo (10k paths) | [MONTE-CARLO-CHOLESKY](/entities/monte-carlo-cholesky-cholesky-multi-asset-correlated-monte-carlo-path-simulator) | Pathwise empirical 97.5% ES | Arbitrary joint copula / gamma | Exotic options & multi-desk aggregation |

## Multi-Asset Covariance Conditioning and Portfolio Risk Attribution

High-dimensional books ($N > T$) suffer singular sample matrices and asynchronous-close negative eigenvalues that crash [Cholesky Multi-Asset Correlated Monte Carlo Path Simulator](/entities/monte-carlo-cholesky-cholesky-multi-asset-correlated-monte-carlo-path-simulator) runs; fix them via [Cholesky PSD matrix repairs](/guides/cholesky-covariance-matrix-positive-definiteness-monte-carlo-var) and [Ledoit-Wolf covariance shrinkage](/guides/ledoit-wolf-covariance-shrinkage-high-dimensional-portfolio-var). For desk capital allocation, [Euler Component VaR attribution](/guides/component-marginal-incremental-var-euler-risk-decomposition) sums additively to 100% of firm VaR, paired with [stressed-beta correlation breakdown math](/guides/fx-equity-correlation-breakdown-stressed-beta-var-amplification).

![Comparison of standardized tail loss multipliers across Gaussian, Student-t, Cornish-Fisher, and EVT GPD models](/api/figure?type=bars&title=Standardized%20Tail%20Loss%20Multipliers%20Under%20Leptokurtic%20Return%20Regimes&subtitle=Unit-volatility%201-day%20loss%20multiples%20comparing%2099%25%20VaR%20and%2097.5%25%20Expected%20Shortfall%20across%20distributional%20assumptions.&steps=Gaussian%20Baseline%3A%3AAssumes%20zero%20skew%20and%20mesokurtic%20kurtosis%20K%3D3.0%7CStudent-t%20%28nu%3D5%29%3A%3APower-law%20heavy%20tails%20scaled%20to%20unit%20variance%7CCornish-Fisher%2099%25%3A%3ACalibrated%20at%20S%3D-0.80%20skew%20and%20Ke%3D4.50%20kurtosis%7CEVT%20GPD%20%28xi%3D0.28%29%3A%3A95th-percentile%20threshold%20excess%20tail%20integral&metrics=GAUSSIAN%2097.5%25%20ES%3A%3A2.34%20sigma%3A%3AZero%20excess%20kurtosis%20baseline%7CSTUDENT-T%2097.5%25%20ES%3A%3A2.89%20sigma%3A%3ASymmetric%20fat-tail%20uplift%7CCORNISH-FISHER%2099%25%3A%3A3.38%20sigma%3A%3ANegative%20skew%20penalty%20included%7CEVT%20GPD%2097.5%25%20ES%3A%3A3.62%20sigma%3A%3AExtreme%20tail%20shape%20xi%3D0.28)

## Regulatory Backtesting, Stressed ES, and Non-Modellable Risk Factors

No quantitative VaR or ES engine survives regulatory audit by [UK FCA](/regulator-index/fca) or [Central Bank of Ireland](/regulator-index/cbi) supervisors without passing statistical backtests on 250-day clean and hypothetical PnL series. Unconditional exception counting relies on [Kupiec POF likelihood-ratio traffic-light zones](/guides/kupiec-pof-likelihood-ratio-var-exception-backtesting-zones), where 0–4 breaches at 99% VaR keep a bank's capital multiplier at $1.50\times$, 5–9 breaches trigger Amber surcharges up to $1.92\times$, and 10+ breaches force disqualification into standardized capital. Because four breaches clustered inside a single volatility week are far more dangerous than four isolated breaches across twelve months, risk validators pair Kupiec with [Christoffersen Markov independence conditional coverage](/guides/christoffersen-markov-independence-conditional-coverage-test). Testing Expected Shortfall directly is trickier because Gneiting proved ES is not standalone elicitable; banks solve this via joint $(VaR, ES)$ scoring and [Acerbi-Szekely Z2 ES backtesting statistics](/guides/acerbi-szekely-expected-shortfall-backtesting-z2-statistic).

Finally, Basel FRTB IMA capital aggregates modeled and unmodeled stress components:
1. **Stressed Expected Shortfall ($SES$)**: Calibrating reduced-factor ES over the most severe 12-month crisis window (e.g., 2008 GFC or March 2020) and scaling by the current full-to-reduced ratio $ES_{F,C}/ES_{R,C}$, detailed in [FRTB Stressed ES 12-month crisis calibration](/guides/stressed-expected-shortfall-frtb-12-month-crisis-window-calibration).
2. **Non-Modellable Risk Factors (NMRFs)**: Any tenor or basis point failing the Risk Factor Eligibility Test (fewer than 24 real committed price observations per year or a 90-day gap without 4 observations) incurs zero-diversification stress add-ons under [Basel NMRF stress scenario capital math](/guides/basel-non-modellable-risk-factors-nmrf-stress-scenario-add-ons).

Explore the complete 20-article research index on the [Parametric VaR & Expected Shortfall Topic Hub](/topics/parametric-var-expected-shortfall), verify institutional clearing models such as [CME SPAN 2 VaR](/entities/cme-span2-var-cme-span-2-historical-stress-value-at-risk-var-model) against broker disclosures in our [broker lookup directory](/broker-lookup/admirals), and audit calculation transparency across our [institutional methodology](/methodology).

## Frequently asked questions

### Why did Basel III FRTB replace 99% Value-at-Risk with 97.5% Expected Shortfall?

Gaussian 99% VaR ignores loss severity beyond the 1% threshold and violates subadditivity, occasionally penalizing portfolio diversification. Calibrating Expected Shortfall at 97.5% preserves the exact normal-distribution capital baseline (2.3378 sigma vs 2.3263 sigma) while penalizing fat-tailed kurtosis and satisfying coherent risk axioms.

### When does square-root-of-time scaling fail in parametric VaR calculators?

Multiplying 1-day VaR by sqrt(10) assumes independent, identically distributed zero-mean returns with constant variance. If daily returns exhibit positive AR(1) autocorrelation (rho = +0.20), true 10-day volatility exceeds sqrt(10) scaling by roughly 20%, whereas GARCH(1,1) mean reversion pulls post-spike multi-day volatility below sqrt(T).

### What causes Cornish-Fisher VaR quantiles to become non-monotonic?

Because Cornish-Fisher truncates an asymptotic Hermite polynomial expansion at skewness and excess kurtosis, extreme moment combinations violate the Maillard domain bound, causing the cubic z-score polynomial to fold backward so 99.5% VaR erroneously prints lower than 99.0% VaR.

### How many VaR breaches trigger a regulatory capital surcharge over 250 trading days?

Under Basel desk-level and bank-wide 99% VaR backtesting over 250 business days, 0 to 4 exceptions keep the model in the Green Zone (1.50x ES multiplier), 5 to 9 exceptions place it in the Amber Zone (1.70x to 1.92x multiplier), and 10 or more breaches trigger Red Zone model disqualification.


## Source URLs

- https://www.bis.org/bcbs/publ/d457.htm
- https://www.bis.org/basel_framework/chapter/MAR/33.htm
- https://www.bankofengland.co.uk/prudential-regulation/publication/2023/november/implementation-of-the-basel-3-1-standards
- https://www.cmegroup.com/clearing/risk-management/span-overview.html

## Citation

- APA: Elena Rostova, FRM, CQF. (2026). Parametric VaR, Expected Shortfall & Fat-Tail Risk Calibration: 2026 Quantitative Guide. QuantCalc Tools. https://quantcalctools.com/guides/parametric-var-expected-shortfall-fat-tail-calibration-guide
- Canonical article: [Parametric VaR, Expected Shortfall & Fat-Tail Risk Calibration: 2026 Quantitative Guide](https://quantcalctools.com/guides/parametric-var-expected-shortfall-fat-tail-calibration-guide)
- Site research index: [QuantCalc Tools](https://quantcalctools.com/llms-full.txt)
