Skip to content

Value at Risk (VaR): Definition

Value at Risk (VaR) answers one question: over a set horizon and at a chosen confidence level, how much could a position lose under normal market conditions? Formally, VaR is the quantile of the loss distribution: the loss that will not be exceeded with probability cc. A one-day 99% VaR of US$5M means that on 99 days out of 100 the loss should stay below US$5M, and on roughly 1 day in 100 it is expected to be worse. VaR always carries two numbers, the confidence level cc and the horizon, and it is a one-tailed statement about the left (loss) tail only.

Try it yourself

Value at Risk — three ways

How much could this book lose on a bad day? VaR draws a line in the loss tail at confidence 95%. The parametric line assumes a normal curve, VaR = (z·σ·√h − μ·h)·V with z = 1.645 (one-tailed). The historical and Monte-Carlo methods read VaR off a simulated loss distribution instead. Under normality the three roughly agree; fat tails pull the historical figure deeper.

1-day 95% parametric VaRUS$3.29M
-6.0%-4.0%-2.0%0.0%2.0%4.0%6.0%Portfolio return over 1 day (%)VaR cutoff5% loss tail
Parametric (normal) · largestUS$3.29M
Historical (fat tails)US$3.05M
Monte-Carlo (5,000 draws)US$3.27M
z (95%, one-tailed) 1.645Horizon σ√h 2.00%Largest VaR Parametric
Portfolio value VUS$100M
Daily volatility σ2.0%
Daily expected return μ+0.00%
Horizon h1 day
Confidence level c
The three estimates land near US$3.29M, but the historical figure of US$3.05M is about the same, since this sample happens not to bite past the Gaussian tail here. The largest of the three is Parametric.
Discuss: which method would you report to a regulator, and which to your own risk committee?

Why it matters

VaR draws a line in the loss tail and says "we do not expect to cross this on an ordinary day." It compresses a whole distribution of outcomes into a single dollar figure a manager can act on. The catch is that it is a threshold, not a worst case: it tells you the edge of the bad region, never how deep the losses run once you fall past it. Quoting a VaR without its confidence level and horizon is meaningless, like a speed with no units.

Formulas

VaR as a loss quantile
Pr ⁣(L>VaRc)=1c\Pr\!\left(L > \mathrm{VaR}_c\right) = 1 - c
The loss LL exceeds VaRc\mathrm{VaR}_c only with probability $1-c$. With c=0.99c = 0.99, losses breach VaR about 1 percent of the time. VaR fixes the probability and reads off the loss threshold.
Loss tail probability (one-tailed)
VaRc=inf{:Pr(L)c}\mathrm{VaR}_c = \inf\{\ell : \Pr(L \le \ell) \ge c\}
VaR is the smallest loss level \ell that the loss stays at or below with probability at least cc. It is a quantile of one tail, so a 99% VaR uses the 1st percentile of the profit-and-loss distribution.

Worked examples

Scenario

A desk reports a one-day 99% VaR of US$5M on a US$100M portfolio. Interpret it, and state what it does and does not promise.

Solution

On about 99 of every 100 trading days the one-day loss should be smaller than US$5M; on roughly 1 day in 100 the loss is expected to equal or exceed US$5M. That breach is 5 percent of capital. VaR does not say how large the loss is on that bad day, only that it lands beyond US$5M. To gauge the depth of the tail you need expected shortfall, not VaR.

Scenario

The same desk also quotes a one-day 95% VaR. Why is the 95% figure smaller than the 99% figure for the same portfolio?

Solution

A higher confidence level pushes further into the loss tail, so it captures a larger loss. The 95% VaR is the 5th-percentile loss, the 99% VaR is the more extreme 1st-percentile loss. For any fixed horizon, raising cc from 0.95 to 0.99 always weakly increases VaR because you are asking about a rarer, deeper loss.

Common mistakes

  • VaR is the maximum possible loss. It is only a threshold for a given confidence level; losses beyond VaR are not just possible but expected on roughly (1c)(1-c) of days.
  • A VaR number stands on its own. VaR is meaningless without both a confidence level and a horizon; "VaR of US$5M" says nothing until you add "one-day, 99%".
  • A higher confidence level makes the portfolio safer. The confidence level is a reporting choice; raising it only raises the quoted VaR number, it does not change the underlying risk.
  • VaR is two-tailed like a confidence interval. VaR looks only at the loss (left) tail, so it uses a one-tailed quantile, not a symmetric interval.

Revision bullets

  • VaR = the loss not exceeded with probability cc over a set horizon
  • Always quote both the confidence level and the horizon
  • One-tailed: a 99% VaR is the 1st-percentile loss
  • It is a threshold, not a worst case or maximum loss
  • Says nothing about loss size beyond the threshold

Quick check

A one-day 99% VaR of US$5M means that, under normal conditions,

Why must a VaR figure always state a confidence level and a horizon?

Connected topics

Sources

  1. Jorion (2007), Ch. 5
    Jorion, P. Value at Risk: The New Benchmark for Managing Financial Risk. 3rd ed. McGraw-Hill, 2007. ISBN 978-0-07-146495-6.
    Defines VaR as a quantile of the loss distribution at a stated confidence level and horizon.
  2. Hull (2018), Ch. 12
    Hull, J. C. Risk Management and Financial Institutions. 5th ed. Wiley, 2018. ISBN 978-1-119-44811-2.
    Introduces VaR, the confidence-level / time-horizon convention, and its one-tailed interpretation.
How to cite this page
Dr. Phil's Quant Lab. (2026). Value at Risk (VaR): Definition. Derivatives Atlas. https://phucnguyenvan.com/concept/frm-var-definition