Value at Risk (VaR) estimates the maximum amount a portfolio could lose over a specific time period at a given confidence level. A 95% one-month VaR of $5,000, for example, means there is a 95% chance your portfolio will not lose more than $5,000 over the next month. Every VaR calculation has three components: a time period, a confidence level, and a loss amount (expressed in dollars or as a percentage).
Three main methods are used to calculate VaR, and each one approaches the problem differently. The right choice depends on the complexity of your portfolio and the tools you have available.
The Parametric (Variance-Covariance) Method
The parametric method is the most straightforward way to calculate VaR. It assumes your portfolio returns follow a normal distribution (the classic bell curve) and uses just a few inputs: the portfolio’s value, its standard deviation (a measure of how much returns typically swing), and a Z-score that corresponds to your chosen confidence level.
The formula for a single asset is:
VaR = Portfolio Value × Z-score × Standard Deviation
Suppose you have a $500,000 portfolio with an annual standard deviation of 7%, and you want to calculate VaR at the 95% confidence level. The Z-score for 95% confidence is 1.645. Plug in the numbers: $500,000 × 1.645 × 0.07 = $57,575. That means, with 95% confidence, you would not expect to lose more than $57,575 over the course of a trading year.
If you want 99% confidence instead, swap the Z-score to 2.326. The higher confidence level produces a larger VaR number because you’re capturing a wider range of possible outcomes.
Portfolios With Multiple Assets
When your portfolio holds two or more securities, you need to account for how those assets move relative to each other. The process starts by calculating the combined portfolio volatility using each asset’s weight, each asset’s standard deviation, and the correlation between them. For two assets, you square each asset’s weight and standard deviation, multiply them together, then add a term that captures the correlation effect: two times the weight of the first asset, times the weight of the second, times their correlation coefficient, times both standard deviations. Take the square root of that sum to get portfolio volatility, then multiply by the Z-score and the portfolio’s total value.
This sounds complex on paper, but it’s easily handled in a spreadsheet. The key data you need for each asset pair is the correlation coefficient, which you can calculate from historical return data or find through financial data providers.
Historical Simulation
Historical simulation skips the normal distribution assumption entirely. Instead, it uses actual past returns to build a picture of what could happen next.
Start by collecting daily (or weekly, or monthly) returns for your portfolio over a historical window, often 250 to 500 trading days. Apply each of those past returns to your current portfolio value to generate a distribution of hypothetical profit and loss outcomes. Then sort those outcomes from worst to best.
For a 95% confidence level, find the return at the 5th percentile of that sorted list. If you have 500 days of data, that’s the 25th worst day. The loss corresponding to that return is your VaR.
The advantage here is simplicity and realism. You don’t need to assume returns follow any particular shape. Fat tails, skewness, and other quirks in real market data are automatically captured. The downside is that the method is entirely backward-looking. If the historical window you chose was unusually calm, your VaR estimate will understate risk. If it included a crisis period, it may overstate it.
Monte Carlo Simulation
Monte Carlo simulation generates thousands (sometimes hundreds of thousands) of random scenarios for how your portfolio might perform, then measures the losses across all those scenarios. It is the most flexible of the three methods because you can model complex instruments like options, structured products, and portfolios with nonlinear payoffs.
The process works like this: you define assumptions about return distributions, volatilities, and correlations for each asset. A computer then randomly generates a large number of possible return paths based on those assumptions. Each path produces a hypothetical portfolio value at the end of your chosen time horizon. You sort the results and find the loss at your confidence threshold, just as in historical simulation.
A Monte Carlo VaR conclusion might read: with 95% confidence, you would not expect to lose more than 15% during any given month. The strength of this approach is its adaptability. The tradeoff is computational cost and the fact that your results are only as good as the assumptions you feed in. Poorly chosen distributions or correlations will produce misleading numbers.
How to Read a VaR Number
A VaR figure always reads as a statement about probability. If a financial firm says an asset has a 3% one-month VaR of 2%, that means there is a 3% chance the asset could decline by 2% or more during the next month. A one-day VaR of $5,000 at the 95% confidence level means there is a 5% chance the investment could lose more than $5,000 in a single day.
Two things to keep in mind when interpreting these numbers. First, the confidence level and time horizon must match. A 95% daily VaR and a 95% annual VaR are answering very different questions. Roughly speaking, you can scale a daily VaR to a longer period by multiplying it by the square root of the number of trading days, though this shortcut works best under the parametric method’s assumptions. Second, VaR tells you the boundary of “normal” losses. It says nothing about how bad things could get beyond that boundary.
What VaR Does Not Tell You
VaR’s biggest blind spot is tail risk, the severity of losses that fall beyond the confidence threshold. A portfolio with a 99% daily VaR of $10 million might lose $11 million on a bad day, or it might lose $500 million. VaR treats both possibilities the same: they’re both in the 1% tail, and VaR ignores them.
This creates a real practical problem. A trader told to keep daily 99% VaR below $10 million could, in theory, construct a portfolio where there’s a 99% chance of staying within that limit and a 1% chance of catastrophic loss. The VaR constraint would appear satisfied, but the actual risk would be enormous.
VaR also lacks a mathematical property called sub-additivity. In plain terms, this means that combining two portfolios can sometimes produce a VaR that is higher than the sum of each portfolio’s individual VaR. That’s counterintuitive, because diversification should reduce risk, not increase it. This flaw means VaR does not always behave consistently when aggregating risks across desks or business units.
Expected Shortfall as a Complement
Expected Shortfall, also called Conditional VaR (CVaR), addresses the tail risk problem directly. While VaR asks “what is the maximum loss at my confidence level?”, Expected Shortfall asks “if losses exceed the VaR threshold, what is the average loss I should expect?”
The calculation uses the same confidence level and time horizon as VaR. For a 95% Expected Shortfall, you take all the outcomes that fall in the worst 5% of the distribution and average them. This gives equal weight to every loss scenario beyond the threshold, rather than focusing on a single cutoff point.
Expected Shortfall is considered a “coherent” risk measure, meaning it satisfies sub-additivity and behaves logically when you combine portfolios. Many risk managers use VaR as a quick summary and Expected Shortfall as the deeper diagnostic, particularly for portfolios exposed to rare but severe events.
Choosing the Right Method
For a portfolio of stocks and bonds with reasonably normal return patterns, the parametric method is fast and easy to implement in a spreadsheet. You need historical return data to estimate standard deviations and correlations, and the math is transparent.
If your portfolio has experienced significant market dislocations and you want those events reflected in the estimate, historical simulation is a better fit. It requires a clean dataset of past returns but no distributional assumptions.
If your portfolio contains options, derivatives, or other instruments whose payoffs are nonlinear, Monte Carlo simulation is typically the only method that captures the risk accurately. It requires more computing power and careful assumption-setting, but it handles complexity that the other two methods cannot.
Whichever approach you use, treat VaR as one tool in a broader risk assessment. It quantifies normal-range losses effectively but stays silent on worst-case scenarios. Pairing it with Expected Shortfall and stress testing gives a more complete picture of what your portfolio could face.