Value at Risk (VaR): Measuring Potential Portfolio Loss

Key Takeaways

• Value at risk gives the minimum loss expected on the worst X% of days; a 95% one-day VaR of $200,000 means losses exceed that amount about 5 trading days per 100.
• VaR is not a worst-case number, it says nothing about how large losses can be on the 5% of days that breach the threshold.
• The parametric (normal distribution) method consistently understates tail risk for equity and credit portfolios with fat-tailed returns.
• VaR's failure to be sub-additive (combining two portfolios can raise total VaR) is one reason Basel III replaced it with expected shortfall for market risk capital.

What It Is

Value at Risk (VaR) is the minimum loss expected on a portfolio over a defined horizon at a defined confidence level, under assumed market conditions. A 95 percent one-day VaR of 1,000,000 USD means that on 95 percent of trading days the portfolio is expected to lose no more than one million dollars in a single day.

VaR is usually quoted in currency units or as a percentage of portfolio value. The two choices you must specify are the time horizon (often one day or ten days) and the confidence level (usually 95 percent or 99 percent).

The Intuition

Standard deviation tells you how dispersed returns are on both sides of the mean. That symmetric view is useful, but a risk manager cares about the left tail, not the right one. A fund does not get stopped out for winning 4 percent in a day. It gets stopped out for losing 4 percent.

VaR compresses the left tail of the return distribution into one clean statement. It shifts the conversation from "this portfolio is volatile" to "on 19 out of 20 days, the loss should be contained inside this number." Boards, regulators, and trading desks all want that kind of clean answer, which is why VaR spread so quickly after JP Morgan published the RiskMetrics methodology in 1994.

How It Works

Three methods dominate in practice. Each estimates the same quantile of the loss distribution using different assumptions.

1. Historical simulation. Take the last N trading days of portfolio returns, sort them from worst to best, and read off the 5th percentile for a 95 percent VaR. No distributional assumption. Good when you trust the historical window to resemble the future.
2. Parametric (variance-covariance). Assume returns are approximately normal, estimate the mean and volatility, and use the standard normal quantile to scale the loss. Fast and analytical, but wrong in the tail when returns are fat-tailed.
3. Monte Carlo simulation. Specify a model of how risk factors behave, draw thousands of random paths, price the portfolio under each, and read the percentile off the simulated P&L distribution. Flexible for options and nonlinear books, but only as honest as the model used to generate the paths.

The parametric formula most textbooks introduce is:

VaR = -(mu - z * sigma) * V

Where mu is the expected return, sigma is the volatility of returns, z is the standard-normal quantile (1.645 for 95 percent, 2.326 for 99 percent), and V is portfolio value. The negative sign flips a loss into a positive VaR number.

VaR scales across horizons under the square-root-of-time rule when returns are independent and identically distributed:

VaR(T) ≈ VaR(1) * sqrt(T)

That approximation breaks down under autocorrelation or volatility clustering, which are common in real markets.

Worked Example

You manage a 10,000,000 USD portfolio with a daily return volatility of 1.2 percent and an assumed daily mean return of zero.

For a 95 percent one-day VaR using the parametric method:

VaR = 1.645 * 0.012 * 10,000,000 = 197,400 USD

So on about 5 trading days out of every 100, the portfolio is expected to lose more than 197,400 USD in a single session. Switching the confidence level to 99 percent raises the quantile to 2.326 and the VaR to roughly 279,120 USD. The number grew, but the question changed: you are now looking at the 1-in-100 day, not the 1-in-20 day.

Run the same portfolio through historical simulation on the last 500 trading days and you might get 230,000 USD instead, because real returns had fatter left tails than the normal distribution assumes.

Frequently Asked Questions

Q: What is value at risk in simple terms? VaR tells you the loss threshold you should expect to stay below on most days. A 95% one-day VaR of $200,000 means that on 19 out of 20 trading days, you expect to lose no more than $200,000.

Q: How does value at risk affect investment decisions? Trading desks use VaR limits to cap how much market risk each desk can carry. When a desk's VaR approaches its limit, it must reduce positions or seek an exemption. It also drives regulatory capital requirements under the Basel framework.

Q: What is a real-world example of value at risk? A $10 million portfolio with 1.2% daily volatility has a parametric 95% one-day VaR of about $197,000. In practice, the historical simulation method on the same portfolio might produce $230,000 because real returns have fatter tails than a normal distribution assumes.

Q: How can investors avoid the biggest VaR mistakes? Always pair VaR with Conditional VaR (expected shortfall) so you know the severity, not just the frequency, of bad days. Use historical simulation alongside parametric VaR to check whether the normal-distribution assumption is distorting the number.

Q: How is value at risk different from expected shortfall? VaR defines the threshold loss at a confidence level and says nothing about losses beyond it. Expected shortfall (CVaR) averages all losses that exceed the VaR threshold, answering how bad it actually gets on the worst days. Basel III uses expected shortfall for market risk capital precisely because it is a more complete tail measure.

Common Mistakes

1. Treating VaR as a worst-case number. It is not. A 95 percent VaR is silent about what happens on the 5 percent of days that breach it. Losses beyond VaR can be many multiples larger, which is the core reason Conditional VaR exists.
2. Assuming normality for fat-tailed assets. Equity indices, credit, and especially options portfolios produce far more extreme losses than the normal distribution predicts. Parametric VaR computed on normal assumptions consistently understates true tail risk, a failure the 2008 crisis made painfully public.
3. Picking a window that hides recent stress. Historical VaR estimated on a calm two-year window will ignore prior crises and look reassuring right until volatility returns. Rolling windows need thought, and stress scenarios should supplement VaR, not replace it.
4. Ignoring non-linearity. Portfolios with options, convertibles, or leveraged products have P&L that curves sharply with the underlying. Parametric and delta-normal VaR miss this curvature entirely. Monte Carlo or full revaluation is the right tool.
5. Forgetting VaR is not sub-additive. Combining two portfolios can produce a total VaR that is larger than the sum of the parts, which violates a basic property a coherent risk measure should have. This is one of the formal arguments that pushed regulators toward expected shortfall in Basel III.
6. Using a single confidence level for every decision. A trading desk, a regulator, and a risk committee each care about different parts of the tail. Report VaR at 95 percent and 99 percent together, and pair both with scenario analysis.

Sources

1. CFA Institute. "Measuring and Managing Market Risk." https://www.cfainstitute.org/en/membership/professional-development/refresher-readings/measuring-managing-market-risk
2. CFA Institute. "VAR: Seductive but Dangerous." Financial Analysts Journal, 1995. https://rpc.cfainstitute.org/research/financial-analysts-journal/1995/var-seductive-but-dangerous
3. Corporate Finance Institute. "Value at Risk." https://corporatefinanceinstitute.com/resources/career-map/sell-side/risk-management/value-at-risk-var/
4. Britannica Money. "Value at Risk (VAR): Meaning, Methods, & How to Calculate." https://www.britannica.com/money/value-at-risk-meaning
5. Basel Committee on Banking Supervision. "Fundamental Review of the Trading Book: A Revised Market Risk Framework." BCBS 265. https://www.bis.org/publ/bcbs265.pdf

Disclaimer

This article is educational content only and is not financial advice. Nothing here is a recommendation to buy, sell, or hold any security. Consult a licensed advisor before making investment decisions.