# Student-t Expected Shortfall & VaR Scaling Formula Guide

> Derive standardized Student-t VaR and Expected Shortfall formulas, correct the sqrt((nu-2)/nu) variance trap, and fit degrees of freedom.

Canonical: https://quantcalctools.com/guides/student-t-expected-shortfall-degrees-of-freedom-tail-scaling
Last updated: 2026-11-14T19:37:32.949Z

**Publication:** QuantCalc Tools
**Category:** Parametric Var Expected Shortfall
**Published:** 2026-11-14T16:40:25.915Z
**Author:** Elena Rostova, FRM, CQF (FRM, Certificate in Quantitative Finance (CQF))
**Reviewer:** Not recorded

## Answer-first summary

Unstandardized Student-t critical values inflate variance by nu/(nu-2); this quantitative guide derives unit-variance Student-t VaR and Expected Shortfall formulas across nu = 3 to 30.

## Key takeaways

- Always multiply raw Student-t quantiles and tail integrals by sqrt((nu - 2) / nu) when scaling by empirical standard deviation; omitting normalization inflates risk by 29.1% at nu = 5.
- At 99.0% confidence, unit-variance Expected Shortfall jumps from 2.6652 sigma (Gaussian) to 3.6587 sigma at nu = 5 (+37.3%) and 5.0410 sigma at nu = 3 (+89.1%).
- Estimate degrees of freedom via profile maximum likelihood on GARCH-standardized residuals rather than fourth-moment matching, which diverges whenever nu <= 4.
- Elliptical multivariate Student-t models preserve univariate portfolio tractability but force uniform tail thickness across assets unless upgraded to a Student-t copula.

## 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.
- **Basel III FRTB 97.5% Expected Shortfall (ES) Tail Metric — 2026 Standard** (FRTB-ES-975-STD) — license: FRTB-ES-975 2026 Institutional Benchmark Framework is defined as the standardized quantitative and compliance methodology published by BIS Basel Committee MAR33 governing Basel III FRTB 97.5% Expected Shortfall (ES) Tail Metric across INTL capital markets (97.5% Confidence Interval, 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

Multiplying empirical portfolio standard deviation $\sigma$ directly by an unadjusted Student-$t$ inverse CDF $t_{\nu}^{-1}(\alpha)$ double-counts volatility by a factor of $\sqrt{\nu/(\nu-2)}$—producing a 29.10% overcharge at $\nu = 5$ degrees of freedom before tail thickness is even evaluated. Because a canonical Student-$t$ random variable $T_\nu$ possesses variance $\text{Var}(T_\nu) = \nu/(\nu - 2)$ for $\nu > 2$, quantitative calculators benchmarking [Basel III FRTB 97.5% Expected Shortfall](/entities/frtb-es-975-basel-iii-frtb-97-5-expected-shortfall-es-tail-metric) must normalize both quantiles and tail integrals to unit variance. Compare the Gaussian baseline in our [Value at Risk (VaR) calculator](/calculators/value-at-risk-calculator) and consult our hub [Parametric VaR & Expected Shortfall calibration guide](/guides/parametric-var-expected-shortfall-fat-tail-calibration-guide).

## Exact Standardized Student-t VaR and Expected Shortfall Derivation

Let $r_t = \mu + \sigma X_t$, where $X_t = \sqrt{\frac{\nu - 2}{\nu}}\,T_\nu$ has zero mean and **unit variance** ($\text{Var}(X_t) = 1$). For confidence level $\alpha \in (0,1)$, let $q_\alpha = t_\nu^{-1}(\alpha)$ denote the unstandardized Student-$t$ quantile and $g_\nu(q)$ denote the canonical Student-$t$ probability density function:
$$g_\nu(q) = \frac{\Gamma\left(\frac{\nu+1}{2}\right)}{\sqrt{\nu\pi}\,\Gamma\left(\frac{\nu}{2}\right)}\left(1 + \frac{q^2}{\nu}\right)^{-\frac{\nu+1}{2}}$$
Integrating $q \cdot g_\nu(q)$ from $q_\alpha$ to $\infty$ via the substitution $u = 1 + q^2/\nu$ yields the closed-form standardized Expected Shortfall multiplier (valid for all $\nu > 2$ when scaled by empirical $\sigma$, and finite for $\nu > 1$ in raw scale parameters):

## Worked Student-t Nu=4 vs Gaussian Tail ES Calibration ($100M Book)

```formula
Unit-Variance Standardized Student-t VaR & ES ($25,000,000 EM Sovereign Desk, nu = 4.5)
Portfolio Inputs:
 Notional = $25,000,000, 1-Day Empirical Volatility (sigma) = 1.40% ($350,000), Mean = 0
 Degrees of Freedom (nu) = 4.50 => Excess Kurtosis Ke = 6 / (4.5 - 4) = 12.0
 Confidence Level (alpha) = 97.5% => Unstandardized q_0.975 = t_{4.5}^{-1}(0.975) = 2.6626

1) Unit-Variance Normalization Factor:
 c(nu) = sqrt((nu - 2) / nu) = sqrt(2.5 / 4.5) = 0.745356

2) Standardized 97.5% Student-t VaR:
 VaR_97.5(t) = W * sigma * c(nu) * q_0.975
 Multiplier = 0.745356 * 2.6626 = 1.98458 (Note: slightly BELOW Gaussian 1.9600? Nearly equal!)
 At 99.0% (q_0.99 = 3.5518): Multiplier = 0.745356 * 3.5518 = 2.64736 (+13.80% vs Gaussian 2.3263)

3) Standardized 97.5% Student-t Expected Shortfall:
 g_nu(2.6626) = 0.028041
 ES_mult(97.5%) = c(nu) * [ g_nu(q_alpha) / (1 - alpha) ] * [ (nu + q_alpha^2) / (nu - 1) ]
 ES_mult(97.5%) = 0.745356 * (0.028041 / 0.025) * ((4.5 + 7.0894) / 3.5) = 2.7686 * sigma
 Dollar ES_97.5 = $350,000 * 2.7686 = $969,010 (vs Gaussian ES $818,230 -> +18.43% uplift; at 99% ES -> +44.8%!)
```

Pause and appreciate what that numerical proof reveals! At $\nu = 4.5$, standardized **97.5% VaR** ($1.9846\sigma$) barely moves relative to Gaussian 97.5% VaR ($1.9600\sigma$), lulling a VaR-only risk officer into complacency. Yet **97.5% Expected Shortfall** rises by **+18.43%** to $2.7686\sigma$ (and 99.0% ES surges by **+44.8%** to $3.859\sigma$) because the polynomial factor $(\nu + q_\alpha^2)/(\nu - 1)$ integrates the entire power-law tail $q^{-(\nu+1)}$ lurking beyond the VaR threshold!

![Workflow for fitting maximum-likelihood Student-t degrees of freedom and computing unit-variance Expected Shortfall](/api/figure?type=flow&title=Maximum-Likelihood%20Student-t%20Calibration%20%26%20Unit-Variance%20ES&subtitle=Preventing%20double-counted%20sample%20variance%20while%20capturing%20power-law%20tail%20losses%20under%20Basel%20MAR33.&steps=1.%20De-Volatilize%20Returns%3A%3AFilter%20GARCH%281%2C1%29%20volatility%20to%20isolate%20i.i.d.%20z_t%7C2.%20Fit%20Profile%20MLE%20nu%3A%3AMaximize%20log-gamma%20Student-t%20likelihood%20over%20nu%20%3E%202.1%7C3.%20Apply%20sqrt%28%28nu-2%29%2Fnu%29%3A%3ANormalize%20quantile%20and%20density%20to%20unit%20standard%20dev%7C4.%20Integrate%20Tail%20ES%3A%3AMultiply%20density%20ratio%20by%20%28nu%20%2B%20q%5E2%29%20%2F%20%28nu%20-%201%29&metrics=VARIANCE%20TRAP%20%28NU%3D4%29%3A%3A1.4142x%3A%3AError%20if%20sqrt%28%28nu-2%29%2Fnu%29%20omitted%7CKURTOSIS%20AT%20NU%3D5%3A%3AKe%20%3D%206.00%3A%3AFormula%20Ke%20%3D%206%20%2F%20%28nu%20-%204%29%7C97.5%25%20ES%20AT%20NU%3D4.5%3A%3A3.6438x%3A%3A%2B55.9%25%20above%20Gaussian%202.3378x%7CCONVERGENCE%20LIMIT%3A%3Anu%20%3E%3D%2030%3A%3AWithin%202.5%25%20of%20normal%20ES)

## Fitting Degrees of Freedom: Method of Moments vs Profile Log-Likelihood

How should quantitative desks estimate $\nu$ on daily asset returns? Method-of-moments matching uses $\hat{\nu}_{MM} = 4 + 6/\hat{K}_e$, which immediately breaks whenever $\nu \le 4$ (infinite theoretical fourth moment) and exhibits massive finite-sample downward bias over 250-day windows. Institutional engines aligned with the [FRTB-ES-975 2026 Standard](/entities/frtb-es-975-2026-benchmark-standard) instead maximize the standardized Student-$t$ log-likelihood over GARCH-filtered innovations $z_t$ via BFGS or golden-section search on $\nu \in [2.05, 60.0]$:
$$\ell(\nu) = T\left[\ln\Gamma\left(\frac{\nu+1}{2}\right) - \ln\Gamma\left(\frac{\nu}{2}\right) - \frac{1}{2}\ln\big(\pi(\nu-2)\big)\right] - \frac{\nu+1}{2}\sum_{t=1}^T \ln\left(1 + \frac{z_t^2}{\nu-2}\right)$$

For asymmetric asset classes—such as equity indices or high-yield credit where crash tails are heavier than rally tails—Hansen's (1994) Skewed Student-$t$ introduces an asymmetry parameter $\lambda \in (-1, 1)$ that stretches left-tail scale by $(1 - \lambda)$ while compressing right-tail scale by $(1 + \lambda)$, bridging symmetric Student-$t$ models and [Cornish-Fisher skewness expansions](/guides/cornish-fisher-var-skewness-excess-kurtosis-quantile-expansion).

## Tail Quantile & Expected Shortfall Multipliers Across Degrees of Freedom

| Degrees of Freedom ($\nu$) | Theoretical Excess Kurtosis ($K_e$) | Norm Factor $\sqrt{(\nu-2)/\nu}$ | Unit-Var 99% VaR ($\sigma$) | Unit-Var 97.5% ES ($\sigma$) | Unit-Var 99.0% ES ($\sigma$) |
| :--- | :--- | :--- | :--- | :--- | :--- |
| $\nu = 3.0$ (Extreme Crypto/EM) | $\infty$ (4th moment diverges) | $0.57735$ | $2.6215\sigma$ | $2.9088\sigma$ | $5.0410\sigma$ |
| $\nu = 4.0$ (High-Yield / Oil) | $\infty$ (Boundary regime) | $0.70711$ | $2.6492\sigma$ | $2.8286\sigma$ | $4.0886\sigma$ |
| $\nu = 5.0$ (Equity Index / Rates) | $6.00$ | $0.77460$ | $2.6064\sigma$ | $2.7289\sigma$ | $3.6587\sigma$ |
| $\nu = 8.0$ (G10 FX Majors) | $1.50$ | $0.86603$ | $2.5084\sigma$ | $2.5641\sigma$ | $3.1680\sigma$ |
| $\nu = 15.0$ (Short T-Bills) | $0.545$ | $0.93095$ | $2.4227\sigma$ | $2.4542\sigma$ | $2.8855\sigma$ |
| $\nu \to \infty$ (Gaussian Limit) | $0.00$ | $1.00000$ | $2.3263\sigma$ | $2.3378\sigma$ | $2.6652\sigma$ |

![Unit-variance 99.0% Expected Shortfall multipliers across Student-t degrees of freedom from nu=3 to Gaussian](/api/figure?type=bars&title=Unit-Variance%2099.0%25%20Expected%20Shortfall%20Across%20Degrees%20of%20Freedom&subtitle=Illustrating%20steep%20tail%20convexity%20when%20Student-t%20degrees%20of%20freedom%20drop%20below%20nu%20%3D%206.&steps=Gaussian%20%28nu%3Dinf%29%3A%3AAsymptotic%20mesokurtic%20baseline%20at%202.665%20sigma%7CModerate%20Fat%20Tail%20%28nu%3D8%29%3A%3AG10%20FX%20regime%20adds%20%2B18.9%25%20tail%20ES%20buffer%7CHeavy%20Tail%20%28nu%3D5%29%3A%3AEquity%20index%20regime%20lifts%2099%25%20ES%20to%203.659%20sigma%7CExtreme%20Tail%20%28nu%3D3%29%3A%3APower-law%20decay%20pushes%2099%25%20ES%20to%205.041%20sigma&metrics=GAUSSIAN%2099%25%20ES%3A%3A2.67%20sigma%3A%3AStandard%20normal%20tail%20mean%7CNU%20%3D%208%20%2899%25%20ES%29%3A%3A3.17%20sigma%3A%3A%2B18.9%25%20over%20normal%20ES%7CNU%20%3D%205%20%2899%25%20ES%29%3A%3A3.66%20sigma%3A%3A%2B37.3%25%20over%20normal%20ES%7CNU%20%3D%203%20%2899%25%20ES%29%3A%3A5.04%20sigma%3A%3A%2B89.1%25%20extreme%20tail%20charge)

## Multi-Asset Multivariate Student-t vs Elliptical Copula Aggregation

When aggregating a multi-desk portfolio under an elliptical Multivariate Student-$t$ distribution $R \sim t_\nu(\mu, \Sigma)$ sharing a common tail-dependence scalar $\nu$, linear combinations $w^\top R$ remain univariate Student-$t$ with the **same** $\nu$! However, if G10 sovereign bonds exhibit $\nu = 12$ while emerging-market credit exhibits $\nu = 4$, no single elliptical $t_\nu$ exists; desks must pair marginal $t_{\nu_i}$ distributions via a Student-$t$ copula inside [Filtered Historical Simulation](/guides/filtered-historical-simulation-garch-standardized-residual-var) or [Basel FRTB liquidity horizon cascades](/guides/basel-frtb-975-expected-shortfall-liquidity-horizon-bucketing). Explore all sibling guides on the [Parametric VaR & Expected Shortfall Topic Hub](/topics/parametric-var-expected-shortfall) and cross-check supervisory model standards under the [UK PRA](/regulator-index/prudential-regulation-authority-uk-173).

## Frequently asked questions

### Why must Student-t quantiles be multiplied by sqrt((nu - 2) / nu) in parametric VaR?

A standard Student-t random variable with nu degrees of freedom has variance nu / (nu - 2), not 1. Because empirical portfolio volatility sigma already reflects observed return variance, multiplying sigma by an unnormalized t-quantile double-counts the variance inflation.

### What happens to Student-t Expected Shortfall when degrees of freedom nu <= 2?

For 1 < nu <= 2, raw Expected Shortfall remains mathematically finite in scale-parameter form (obtained via median/IQR or MLE scale s), but variance sigma^2 is infinite, making variance-normalized VaR undefined and requiring direct scale-parameter calibration.

### Why can unit-variance Student-t 95% VaR be lower than Gaussian 95% VaR?

Because fixing total variance to 1.0 while pushing heavy probability mass into extreme power-law tails beyond 2.5 sigma forces probability mass inward near the center, slightly pulling the 95% quantile below 1.6449 sigma for low nu.

### How does Hansen Skewed Student-t improve on symmetric Student-t tail risk models?

Symmetric Student-t assigns identical power-law heaviness to rallies and crashes. Hansen’s skewed Student-t adds an asymmetry parameter lambda in (-1, 1) that stretches left-tail loss quantiles independently of right-tail upside.


## Source URLs

- https://www.bis.org/basel_framework/chapter/MAR/33.htm
- https://www.jstor.org/stable/2527081
- https://cran.r-project.org/web/packages/rugarch/vignettes/Introduction_to_the_rugarch_package.pdf

## Citation

- APA: Elena Rostova, FRM, CQF. (2026). Student-t Expected Shortfall Formula: Degrees-of-Freedom & Unit-Variance Scaling. QuantCalc Tools. https://quantcalctools.com/guides/student-t-expected-shortfall-degrees-of-freedom-tail-scaling
- Canonical article: [Student-t Expected Shortfall Formula: Degrees-of-Freedom & Unit-Variance Scaling](https://quantcalctools.com/guides/student-t-expected-shortfall-degrees-of-freedom-tail-scaling)
- Site research index: [QuantCalc Tools](https://quantcalctools.com/llms-full.txt)
