Every investor, trader, and risk manager eventually faces the same uncomfortable question: how much could we actually lose? Value at Risk, or VaR, is one of the most widely used tools for answering it. As the Corporate Finance Institute (CFI) explains, VaR is a statistical measure that quantifies the level of financial risk within a firm, portfolio, or financial position over a specified time frame.
## What VaR Actually Measures
At its core, VaR estimates the maximum potential loss expected over a given period at a chosen confidence level. It rests on three essential inputs: the confidence level (such as 95% or 99%), the time horizon (a day, a month, or longer), and the loss amount.
A classic example makes this concrete. If a portfolio has a one-day 95% VaR of $1 million, there is a 5% probability that the portfolio will lose more than $1 million in a single day — or, put another way, a 95% confidence that the loss will not exceed $1 million.
That single number is powerful because it distills a complex web of positions and risk factors into something decision-makers can grasp at a glance. It is no surprise that VaR is widely used in market risk management, regulatory capital calculation, risk-limit setting, capital allocation, and performance evaluation.
## Three Ways to Calculate VaR
Not all VaR is created equal. CFI outlines three main calculation methods, each with its own trade-offs.
**Historical simulation** uses actual historical market data to revalue a portfolio under past market scenarios. It is non-parametric, meaning it does not assume a particular distribution — but it depends heavily on the historical sample chosen.
**The variance-covariance (parametric) method** assumes returns are normally distributed and relies on standard deviation, correlations, and z-scores. It is relatively simple to apply, but it can understate tail risk when returns stray from a normal pattern.
**Monte Carlo simulation** generates many random scenarios based on specified statistical distributions and correlations. This flexibility makes it well suited to complex or non-linear portfolios, though it can be computationally intensive.
For the parametric approach, a basic formula captures the idea: **VaR = portfolio value × z-score × standard deviation**. When expected return is included, it becomes: **VaR = portfolio value × (mean return − z-score × standard deviation)**.
## The Strengths — and the Blind Spots
The appeal of VaR is easy to understand. It delivers a single, easily understood risk number, aggregates multiple positions and risk factors, enables comparison across portfolios, and supports both regulatory and internal risk limits.
But VaR comes with important limitations. It does not indicate the severity of losses beyond the VaR threshold — the so-called tail risk. It relies on historical data and model assumptions. The normal-distribution assumption may miss fat tails and extreme events, and correlations can shift during periods of market stress. Model risk is always present, and VaR is not always subadditive, which means diversification does not always reduce measured VaR.
Most importantly, VaR is not a worst-case loss measure. It tells you the threshold, not what lies beyond it. For that reason, it is typically complemented by stress testing and backtesting, which help address its limitations.
## The Takeaway
Value at Risk remains a cornerstone of modern risk management — a compact, comparable figure that helps institutions set limits and allocate capital. Yet its value depends on understanding what it can and cannot do. Used thoughtfully and alongside stress testing and backtesting, VaR gives risk managers a clearer view of the dangers lurking in their portfolios. Used blindly, it can offer false comfort about the extremes it was never designed to capture.