Skip to main content
v2026.11,986 entries · CC-BY 4.0

Value at Risk: Formula, Worked Example & Guide

A guide to Value at Risk (VaR): mathematical formula, worked arithmetic, Basel backtesting thresholds, APA reporting standards, and calculation snippets.

Written and maintained by CASRAI Editorial Board

Published · Last updated

On this page

Definition in one paragraph

Value at Risk (VaR) is a quantitative risk metric that estimates the maximum expected loss of a portfolio over a specified time horizon at a given statistical confidence level under normal market conditions. Rather than assessing average returns, VaR quantifies tail risk by setting a threshold loss breached only with a predetermined probability (such as 5% or 1%). Governed under the Basel Committee on Banking Supervision (BCBS) Framework and model risk standards, VaR converts complex portfolio volatility into an intuitive dollar figure or percentage loss.

Formula with every symbol explained

Under the standard parametric (variance-covariance) model, asset returns follow a normal distribution. For significance level α and initial wealth V0, parametric Value at Risk is formulated as:

VaR = (Z × σ – μ) × V0

Where each symbol represents:

  • VaR: The Value at Risk in currency units (or percentage return: VaR% = Z × σ – μ).
  • V0: The initial market value of the portfolio.
  • α (Alpha): The significance level (tail probability). Common levels are α = 0.05 for a 95% confidence interval coverage level, and α = 0.01 for 99% coverage.
  • Z: The standard normal quantile for one-tailed confidence level (1 – α). For α = 0.05, Z ≈ 1.645; for α = 0.01, Z ≈ 2.326. Researchers verify exact values by consulting how to find p-value from z-score.
  • σ (Sigma): The return standard deviation (volatility) over the holding horizon. In multi-asset portfolios, σ is derived from asset weights and the complete covariance matrix: σp = √(wT Σ w).
  • μ (Mu): The expected return over the horizon. For daily horizons, μ is conventionally set to zero (μ = 0) because daily drift is negligible relative to volatility.

For an extended horizon T with independent and identically distributed returns, analysts scale 1-day risk via the square-root-of-time rule: VaRT = VaR1 × √T.

Worked example with real numbers and the arithmetic shown

Consider an institutional portfolio evaluated over a 1-day holding horizon:

  • Portfolio Value (V0): $1,000,000
  • Holding Horizon (T): 1 trading day
  • Daily Expected Return (μ): 0.05% (0.0005)
  • Daily Volatility (σ): 1.50% (0.0150)
  • Critical Values: Z0.95 = 1.644854 (95% confidence); Z0.99 = 2.326348 (99% confidence)

Step 1: Baseline Zero-Drift VaR (μ = 0)

Assuming zero daily drift:

VaR95%, 1-day = Z0.95 × σ × V0 = 1.644854 × 0.0150 × $1,000,000 = $24,672.81

VaR99%, 1-day = Z0.99 × σ × V0 = 2.326348 × 0.0150 × $1,000,000 = $34,895.22

Step 2: Absolute VaR with Expected Drift (μ = 0.0005)

Incorporating positive drift reduces the loss threshold:

For 95% confidence: Loss factor = (1.644854 × 0.0150) – 0.0005 = 0.0241728; VaR95%, absolute = 0.0241728 × $1,000,000 = $24,172.80

For 99% confidence: Loss factor = (2.326348 × 0.0150) – 0.0005 = 0.0343952; VaR99%, absolute = 0.0343952 × $1,000,000 = $34,395.22

Step 3: Multi-Day Horizon Scaling (10-Day Horizon)

Applying the square-root-of-time rule for 10 days (√10 ≈ 3.162278):

VaR99%, 10-day = $34,395.22 × 3.162278 = $108,767.25

Interpretation and reading guidance

A 1-day 99% VaR of $34,395 means that on 99 out of 100 trading sessions, daily portfolio losses should not exceed $34,395 under normal market conditions. Conversely, losses equal or exceed $34,395 on 1% of trading days (roughly 2.5 sessions per 250 trading days).

Estimation Method Distribution Assumption Non-Linear Payoffs Primary Application
Parametric Normal or Student-t Linear delta approximation Liquid linear books, intraday tracking
Historical Simulation Empirical distribution Full revaluation supported Banking portfolios using nonparametric statistics
Monte Carlo Simulation Stochastic diffusion paths Full revaluation supported Exotic derivatives, non-linear portfolios

Under BCBS MAR32, 99% 1-day VaR is backtested against realized daily trading P&L over 250 trading days:

  • Green Zone (0 to 4 exceptions): Model accepted; no capital penalty (plus-factor = 0.00).
  • Yellow Zone (5 to 9 exceptions): Model flagged; capital multiplier increases (plus-factors from 0.40 to 0.85).
  • Red Zone (10 or more exceptions): Model rejected; internal model approval revoked.

Assumptions and common mistakes

Effective risk governance requires understanding model boundaries and failure points:

Core Mathematical Assumptions

  • Normality: Parametric VaR assumes bell-curve returns. Financial assets display fat tails and skewness, causing normal models to understate crash risk.
  • Return Independence: Square-root scaling (√T) requires independent, identically distributed returns. Autocorrelation and volatility clustering distort multi-day projections.
  • Liquidity: Standard models assume positions exit at current prices without market impact, an assumption that fails during liquidity stress.

Violation of Subadditivity

Value at Risk is not a coherent risk measure because it violates subadditivity: VaR(X + Y) ≤ VaR(X) + VaR(Y). For skewed or heavy-tailed portfolios, combined VaR can exceed the sum of individual VaRs. Consequently, Basel III (FRTB) transitioned capital standards to 97.5% Expected Shortfall (ES), which satisfies subadditivity and measures tail loss severity.

Common Practical Mistakes

  • Treating VaR as Worst-Case Loss: VaR marks the start of the loss tail; it does not measure loss depth once breached.
  • Flawed Backtesting: Violating Federal Reserve SR 11-7 by comparing VaR to accounting revenue rather than clean Hypothetical and Actual trading P&L.
  • Overlooking Multivariate Dynamics: Relying on simple linear correlation instead of examining joint tails through marginal distribution vs conditional distribution structures.

How to report it (APA example)

Report parameters, distribution models, horizons, and backtesting outcomes following American Psychological Association (APA 7th edition) standards.

APA Narrative Reporting Example

“Daily portfolio risk was estimated using a 1-day parametric variance-covariance Value at Risk (VaR) model for an institutional equity portfolio (V0 = $1,000,000, daily μ = .0005, daily σ = .0150). At the 99% confidence level (α = .01, Z = 2.33), the 1-day VaR was $34,395 (3.44% of portfolio value), indicating that daily trading losses are expected to exceed this threshold on no more than 1% of trading sessions under stationary market conditions. Over a 250-day rolling backtesting window evaluated against hypothetical trading outcomes, the model recorded three exceptions (1.20% rate), placing it in the Basel Green Zone (p = .24, binomial test) and confirming supervisory adequacy under BCBS MAR32 standards.”

How to calculate it in Excel, R or Python

The snippets below implement 1-day parametric Value at Risk for a $1,000,000 portfolio at 99% confidence with daily volatility σ = 0.015 and expected return μ = 0.0005.

1. Microsoft Excel

# Parametric VaR with expected daily return
= (NORM.S.INV(0.99) * 0.015 - 0.0005) * 1000000

# Parametric VaR with zero drift (mu = 0)
= NORM.S.INV(0.99) * 0.015 * 1000000

# Nonparametric Historical Simulation VaR (returns in column A, rows 2:251)
= -PERCENTILE.INC(A2:A251, 0.01) * 1000000

2. R

# Portfolio parameters
v0 <- 1000000      # Portfolio value in USD
mu <- 0.0005       # Daily mean return
sigma <- 0.015     # Daily return standard deviation
alpha <- 0.01      # Significance level (99% confidence)

# 1-Day Parametric Value at Risk
z_score <- qnorm(1 - alpha)
var_parametric <- (z_score * sigma - mu) * v0
cat(sprintf("1-day 99%% Parametric VaR: $%.2f
", var_parametric))

3. Python

import statistics

# Portfolio parameters
v0 = 1_000_000     # Portfolio value in USD
mu = 0.0005        # Daily mean return
sigma = 0.015      # Daily return standard deviation
conf_level = 0.99  # Statistical confidence level

# Parametric VaR using standard normal quantile
dist = statistics.NormalDist(mu=0.0, sigma=1.0)
z_score = dist.inv_cdf(conf_level)
var_parametric = (z_score * sigma - mu) * v0

print(f"1-Day 99% Parametric VaR: ${var_parametric:,.2f}")

FAQ

What is the difference between Value at Risk (VaR) and Expected Shortfall (ES)?

VaR identifies the loss threshold at a chosen confidence level. In contrast, Expected Shortfall (conditional VaR) calculates the expected average loss given that the VaR threshold has been breached, capturing tail severity beyond the quantile cutoff.

Why did the Basel Framework transition from VaR to Expected Shortfall?

Under Basel III (FRTB), regulators transitioned market risk internal models capital from 99% VaR to 97.5% Expected Shortfall. This resolved VaR's subadditivity violation, where portfolio diversification failed under heavy tails, and captured tail loss severity.

Can Value at Risk be negative?

VaR is reported as a positive value representing capital at risk. However, if an asset has high expected positive drift relative to volatility such that μ > Z × σ, the raw loss formula yields a negative figure, indicating an expected gain even at the tail threshold.

How does the holding period affect Value at Risk?

The holding period reflects the time needed to orderly liquidate positions. Under independent returns, multi-day risk scales with the square root of time (VaRT = VaR1 × √T). Autocorrelation or volatility clustering makes this scaling understate true multi-day risk.

What constitutes an exception in VaR backtesting?

An exception occurs when realized trading losses exceed the model's predicted VaR threshold for that day. Supervisors track exceptions over 250 days to place models into Basel Green (0–4 exceptions), Yellow (5–9 exceptions), or Red (10+ exceptions) zones.

Sources

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about Value at Risk: Formula, Worked Example & Guide

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →