Back to Risk Metrics
Risk Metrics

Value at Risk (VaR)

A statistical measure that estimates the maximum potential loss of a portfolio over a given time period at a specified confidence level. VaR answers the question: "What's the worst I should expect to lose?"

Quick Summary

  • Formula: VaR = Portfolio Value x z-score x σ x t (parametric method)
  • Measures: Maximum expected loss at a given confidence level over a time period
  • Key numbers: "95% daily VaR of $5,000" means: 95% of the time, daily losses won't exceed $5,000
  • Key insight: VaR doesn't tell you what happens in the worst 5% — that's where Conditional VaR comes in

What Is Value at Risk?

VaR became the industry standard risk measure after JP Morgan published its RiskMetrics framework in the 1990s. Banks, hedge funds, and regulators use it daily to quantify market risk. In plain English: if you have a 95% daily VaR of $5,000, it means on 19 out of 20 trading days, your portfolio's loss will not exceed $5,000. On that 20th day (the 5% tail), losses could be anything — $6,000, $20,000, or more. This is both VaR's strength (simple, intuitive) and its limitation (it doesn't describe tail risk).

VaR has three components: a confidence level (typically 95% or 99%), a time horizon (1 day, 1 week, 1 month), and the loss amount in dollars or percentage.

VaR relies on the concept of standard deviation to quantify how returns vary, and complements backward-looking measures like maximum drawdown by providing a forward-looking estimate of potential losses.

How to Calculate VaR: Three Methods

Method 1: Historical VaR

The simplest approach — use actual past returns to estimate future risk:

Step 1: Collect actual past returns (e.g., 500 daily returns)

Step 2: Sort from worst to best

Step 3: At 95% confidence, VaR = the 25th worst return (5% of 500)

Example: $100K portfolio, sorted returns, 25th worst day was -2.1%, so 95% daily VaR = $2,100

Pro: No distribution assumptions. Con: Assumes the past repeats.

Method 2: Parametric (Variance-Covariance) VaR

VaR = Portfolio Value x z x σ x t

Where z = 1.65 for 95% confidence, 2.33 for 99%, σ is the portfolio standard deviation, and t is the time period in days.

Example: $100K portfolio, daily σ = 1.2%

95% 1-day VaR = $100,000 x 1.65 x 0.012 = $1,980

Scale to other periods: multiply by t

10-day VaR = $1,980 x 10 = $6,261

Pro: Fast, easy to compute. Con: Assumes normal distribution (underestimates tail risk).

Method 3: Monte Carlo VaR

The most flexible approach — simulate thousands of possible futures:

Step 1: Simulate thousands (10,000+) of possible future return scenarios

Step 2: Use historical volatility and correlations as inputs

Step 3: Sort simulated outcomes, find the cutoff at the desired confidence level

Pro: Handles any distribution, non-linear instruments, correlations. Con: Computationally intensive, model-dependent.

Method Comparison

FeatureHistoricalParametricMonte Carlo
AssumptionsNone (uses actual data)Normal distributionModel-dependent
Tail riskCaptures if in sampleUnderestimatesFlexible
SpeedFastVery fastSlow
Best forSimple portfoliosQuick estimatesComplex / non-linear
Data neededLong historyMean, std dev, correlationsParameters + simulation engine

How to Interpret VaR

95% VaRStandard corporate risk management — Losses will exceed this approximately 1 day per month (~1 in 20 trading days). The most common confidence level for general portfolio risk monitoring.
99% VaRBank regulatory standard (Basel III) — Losses will exceed this approximately 2-3 days per year (~1 in 100 trading days). Used by banks for capital requirements.
Key caveatVaR says nothing about HOW BAD the tail can be — A 99% VaR of $10K doesn't mean $10K is the worst case — it's the boundary of the normal zone. Losses beyond VaR could be far larger.

VaR vs Maximum Drawdown

Both measure risk, but from very different angles. Use them together for a complete picture:

FeatureValue at RiskMaximum Drawdown
Time frameShort (daily/weekly)Full investment period
DirectionForward-looking estimateBackward-looking fact
What it measuresExpected loss at confidence levelActual worst peak-to-trough loss
Distribution assumptionYes (parametric) or No (historical)None
Tail riskDoesn't measure beyond thresholdCaptures actual worst case
Best forDay-to-day risk monitoringEvaluating total investment experience

Use both: VaR for daily risk management, max drawdown for evaluating overall strategy quality. Together they answer different but equally important questions about your portfolio's risk.

Conditional VaR (Expected Shortfall)

CVaR (also called Expected Shortfall) answers a question VaR cannot: "When losses exceed VaR, how bad is it on average?"

95% VaR = $5,000

On the worst 5% of days, losses exceed $5,000 — but by how much? VaR doesn't say.

95% CVaR = $8,000

When you do have a bad day (the worst 5%), you lose $8,000 on average. CVaR tells the rest of the story.

CVaR is always greater than or equal to VaR. The Basel III framework requires banks to use Expected Shortfall instead of VaR because ES captures the severity of tail losses, not just their probability.

For individual investors, CVaR gives a more honest picture of tail risk. If your 95% VaR is $5,000 but your CVaR is $15,000, the tail is very heavy — and you should plan accordingly.

Real-World Example: $200,000 Diversified Portfolio

Consider a $200,000 diversified portfolio and its VaR using the parametric method:

Asset ClassWeightAnnual σ
Stocks60%16%
Bonds30%5%
Gold10%15%

Portfolio σ (accounting for correlations): 10%

Daily σ: 10% / 252 0.63%

95% daily VaR:

$200,000 x 1.65 x 0.0063 = $2,079

99% daily VaR:

$200,000 x 2.33 x 0.0063 = $2,936

Meaning: on 95% of trading days, this portfolio loses less than ~$2,100. On the worst 1 in 100 days, losses could exceed ~$2,900. These numbers help set realistic expectations for daily portfolio fluctuations.

How Portfolio Genius Calculates VaR

Portfolio Genius automatically calculates Value at Risk for every portfolio, giving you instant visibility into your potential downside:

  • Automated VaR calculation — VaR computed for every portfolio using historical return data, updated as positions change
  • Multiple confidence levels — View both 95% and 99% VaR to understand your risk at different thresholds
  • AI-powered risk alerts — Get notified when VaR exceeds your risk tolerance, with recommendations for reducing exposure
  • Complete risk dashboard — VaR displayed alongside Sharpe ratio, beta, and maximum drawdown for comprehensive risk analysis

Understanding your portfolio's VaR helps you answer a critical daily question: how much could I lose today, and am I comfortable with that number?

Common Mistakes to Avoid

  • Treating VaR as the worst case — It's NOT the worst case; it's the boundary of the normal zone. A 95% VaR of $5,000 means losses exceed $5,000 about once a month — and on those days, the actual loss could be $10,000, $20,000, or more.
  • Ignoring VaR's distribution assumptions — Parametric VaR assumes returns follow a normal distribution. Real markets have fat tails — extreme events occur far more often than the bell curve predicts. This means parametric VaR systematically underestimates tail risk.
  • Using VaR alone without Conditional VaR — VaR tells you where the cliff is; CVaR tells you how far you fall. A portfolio can have a low VaR but catastrophic CVaR if tail losses are extreme. Always look at both numbers together.
  • Comparing VaR across different confidence levels — 95% VaR and 99% VaR are very different numbers. A portfolio with 95% VaR of $5,000 might have a 99% VaR of $8,000+. Always ensure you're comparing the same confidence level and time horizon.
  • Not scaling VaR correctly — To convert daily VaR to t-day VaR, multiply by t, not by t directly. A daily VaR of $1,000 becomes a 10-day VaR of $3,162 (10 3.162), not $10,000. The square root scaling reflects how volatility grows slower than time.

Frequently Asked Questions

What does '95% daily VaR of $5,000' mean?
It means there's a 95% probability that your portfolio won't lose more than $5,000 in a single day. Equivalently, about 1 in 20 trading days (roughly once per month), losses could exceed $5,000. It does NOT mean $5,000 is the maximum possible loss.
Which VaR method is best?
Historical VaR is simplest and makes no distribution assumptions, making it good for straightforward portfolios. Parametric VaR is fastest but assumes normal returns. Monte Carlo VaR is most flexible and handles complex instruments (options, structured products) best. For most individual investors, historical VaR provides the most honest estimate.
What is Conditional VaR (Expected Shortfall)?
CVaR is the average loss when losses exceed the VaR threshold. If 95% VaR is $5,000, CVaR might be $8,000 — meaning when you do have a bad day (the worst 5%), the average loss is $8,000. CVaR is considered a better risk measure because it captures tail severity, not just tail probability.
How is VaR different from standard deviation?
Standard deviation measures all volatility (up and down) equally. VaR translates volatility into a specific dollar loss at a specific confidence level, making it more practical for risk management. VaR uses standard deviation in its parametric calculation (VaR = z × σ × portfolio value) but presents the result in more actionable terms.
Can VaR be used for individual stocks?
Yes, though it's most commonly used at the portfolio level where diversification effects matter. For individual stocks, VaR can help you understand position-level risk. But be cautious: individual stocks often have non-normal return distributions (fat tails), so parametric VaR may significantly underestimate risk.
Why do regulators require Expected Shortfall instead of VaR?
After the 2008 financial crisis, regulators realized VaR gave a false sense of security. A bank could report low VaR while having enormous tail risk. The Basel III framework switched from VaR to Expected Shortfall because ES measures the actual severity of losses beyond the threshold, not just the probability of exceeding it. For individual investors, monitoring both VaR and max drawdown provides a similar level of insight.

Monitor Your Portfolio's Value at Risk

Portfolio Genius calculates VaR and other risk metrics automatically. Know your potential downside before the market moves — not after.