Value at Risk (VaR): Loss Thresholds, Confidence Levels, and Limits

What is Value at Risk?

Value at Risk, usually abbreviated VaR, estimates a loss threshold for a portfolio over a specified time horizon at a chosen confidence level under stated assumptions.

A VaR statement is incomplete unless it includes at least:

1loss amount or percentage
2confidence level
3time horizon
4methodology or assumptions

For example:

11-day 95% VaR = $1 million

can be interpreted, under the model, as a loss threshold that should be exceeded on roughly 5% of days if the modeled return process describes reality.

It does not mean the portfolio cannot lose more than $1 million.

A 95% VaR example

Suppose a portfolio has a one-day 95% VaR of 2%.

The intended interpretation is approximately:

195% of modeled one-day outcomes lose less than 2%
25% of modeled one-day outcomes lose more than 2%

The statistic identifies a percentile of the loss distribution.

It does not tell you whether the outcomes in that worst 5% are losses of 2.1%, 5%, 20%, or more.

That distinction is the central limitation investors should remember.

VaR requires a horizon and confidence level

VaR has no complete meaning without a horizon and confidence level.

A one-day 95% VaR is a different object from:

• one-day 99% VaR;
• ten-day 99% VaR; or
• one-month 95% VaR.

Increasing the confidence level moves farther into the loss tail. Increasing the horizon changes the distribution of possible portfolio outcomes and may introduce compounding, changing positions, liquidity, and serial dependence.

Comparing two VaR numbers without aligning those choices can be misleading.

VaR does not tell you how bad losses can be beyond the cutoff

This is the key analytical boundary:

VaR does not tell you how bad losses can be beyond the cutoff.

Two portfolios can have the same 99% VaR while having radically different outcomes in the worst 1% of cases.

One might experience losses only slightly beyond the VaR threshold. Another might have rare but catastrophic losses.

Expected Shortfall was designed to answer a related tail-severity question by averaging losses beyond a selected tail threshold.

VaR is not maximum possible loss

The word "risk" in the name can make VaR sound more absolute than it is.

VaR is a percentile-based model estimate, not a bound on what can happen.

A position can lose substantially more than its reported VaR. In some instruments, the maximum possible loss can be the entire investment. Leveraged or short positions can create even more complex loss profiles.

A useful risk report should never translate:

199% VaR = $5 million

into:

1maximum loss = $5 million

Those statements mean different things.

Three common VaR methods

CFA Institute's market-risk curriculum describes three widely used approaches: parametric VaR, historical simulation, and Monte Carlo simulation.

Parametric VaR

A parametric method assumes a particular distributional structure and estimates VaR from parameters such as expected returns, variances, and covariances.

A simple normal-distribution approach can be fast and transparent, but it can understate risk when returns are skewed, fat-tailed, or nonlinear.

Historical simulation

Historical simulation applies observed historical market moves to the current portfolio or uses an empirical return distribution.

It avoids assuming a normal distribution, but it is limited by the events contained in the selected history. A crisis that never occurred in the sample cannot appear spontaneously in a simple historical simulation.

Monte Carlo simulation

Monte Carlo VaR simulates many possible market paths from a specified stochastic model.

It can handle complex portfolios and nonlinear payoffs, but the output depends on the model, parameters, dependencies, and scenarios used to generate those paths.

More computation does not remove model risk.

Nonlinear portfolios can challenge simple VaR

Options and other derivatives can have payoffs that respond nonlinearly to market moves.

A simple covariance-based normal model can approximate small changes poorly when delta, gamma, volatility, or other risk factors move materially.

This is one reason a parametric VaR number that looks adequate for a linear stock-and-bond portfolio may be less reliable for a portfolio with substantial optionality.

Risk measurement should reflect the actual payoff structure.

Correlation assumptions matter

Portfolio VaR depends on how positions are expected to move together.

If a model uses historical Correlation or Covariance, diversification can reduce estimated VaR.

During stress, correlations can rise and liquidity can deteriorate at the same time. The diversification benefit embedded in the model can then prove too optimistic.

A serious VaR review should ask what happens when correlation assumptions worsen.

Backtesting checks the model, not the future

Institutions can compare realized daily losses with prior VaR forecasts.

If a 99% daily VaR model were perfectly calibrated under stable conditions, losses beyond the threshold should be rare. Too many exceedances can signal model problems.

But a good historical backtest does not prove the model will remain accurate. Market structure, volatility, portfolio composition, and dependencies can change.

Backtesting is evidence about past calibration, not a guarantee about future tail behavior.

VaR can encourage false precision

A report might publish:

199% 1-day VaR = $4,723,188

The final dollars can create an impression of accuracy that the underlying assumptions do not deserve.

The estimate may depend on noisy volatilities, correlations, distributional choices, and stale positions.

Risk managers should distinguish computational precision from economic certainty.

VaR and Expected Shortfall answer different questions

VaR asks approximately:

1Where is the selected loss cutoff?

Expected Shortfall asks approximately:

1If we are already beyond that cutoff, how large are losses on average?

Expected shortfall therefore contains information about tail severity that VaR omits.

This difference helped motivate the Basel market-risk framework's shift from VaR toward expected shortfall for internal-model capital calculations under stress.

That regulatory choice does not make expected shortfall assumption-free. It reflects a different risk question.

VaR is not drawdown

Maximum Drawdown measures the worst observed peak-to-trough decline in a cumulative wealth path.

VaR measures a modeled or estimated loss percentile over a stated horizon.

A portfolio can have modest one-day VaR yet suffer a large multi-month drawdown through repeated smaller losses.

Conversely, a portfolio can have high daily VaR but recover quickly enough to avoid an extreme long-horizon drawdown.

The metrics should not be substituted for one another.

Liquidity can make realized loss worse

A VaR calculation based on quoted market prices may assume positions can be exited or hedged at those prices.

During stress, bid-ask spreads can widen, market depth can disappear, and large trades can move prices.

Realized loss can therefore exceed the market-move component estimated by a simple VaR model.

The Basel market-risk framework explicitly incorporates liquidity horizons because exit and hedging assumptions matter in stressed markets.

How investors should read a VaR number

Before relying on VaR, ask:

1. What is the confidence level?
2. What is the time horizon?
3. Is the result stated in dollars or percentage terms?
4. Is the method parametric, historical simulation, or Monte Carlo?
5. What distribution and correlation assumptions are used?
6. Does the portfolio contain nonlinear derivatives?
7. What historical period or scenarios drive the estimate?
8. How many VaR exceedances have occurred in backtesting?
9. What does expected shortfall say about losses beyond the cutoff?
10. What liquidity, leverage, and drawdown risks are omitted?

Grizzly Bulls' Models can be reviewed with these tail-risk questions without turning this article into live VaR authority. The Macroeconomic Conditions Index provides separate stress-regime context, but it does not set the VaR confidence level, horizon, or loss distribution.

Sources and further reading

• CFA Institute, 2026, Measuring and Managing Market Risk: https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/measuring-managing-market-risk
• CFA Institute, 2026, Portfolio Mathematics: https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/portfolio-mathematics
• Bank for International Settlements, Market risk terminology: https://www.bis.org/committees/bcbs/basel-framework/standard/mar/10/inforce/2023-01-01/published/2020-03-27
• Bank for International Settlements, Revised market risk framework executive summary: https://www.bis.org/publications/fsi-summary-revised-market-risk-framework-executive-summary

Explore more topics in the Financial Research Encyclopedia.