Value at Risk (VaR) estimates the maximum loss a portfolio could suffer over a set time horizon at a given confidence level — e.g., a 1-day 95% VaR of $2M means a 5% chance of losing more than $2M in a single day under normal market conditions.
Why It Matters
VaR is the risk metric regulators and boards actually ask for: banks use it to set trading limits and calculate regulatory capital under Basel market-risk rules, and asset managers report it to clients as a single, comparable number across very different portfolios. A firm that can't produce a defensible VaR number struggles to get a trading desk approved or a fund past institutional due diligence.
How It Works in Practice
- 1Choose a method: historical simulation (replay actual past returns), variance-covariance (assume normally distributed returns), or Monte Carlo (simulate thousands of random price paths)
- 2Set the confidence level (95% and 99% are standard) and time horizon (1-day is common for trading desks, 10-day for regulatory capital)
- 3Aggregate position-level risk into a single portfolio number, accounting for correlations between holdings
- 4Backtest continuously — count how often actual losses exceed the VaR estimate and compare to the expected exception rate
Common Pitfalls
VaR says nothing about the size of losses beyond the threshold — a 95% VaR ignores exactly the tail events that cause real damage, which is why regulators now also require Expected Shortfall (CVaR)
Historical-simulation VaR is only as good as the lookback window; a calm 2-year window will understate risk right before a crisis
Correlations that hold in normal markets often break down in a crisis, understating portfolio-level VaR exactly when it matters most
Key Metrics
| Metric | Target | Formula |
|---|---|---|
| 1-day 95% VaR | Firm risk-limit dependent | Portfolio value x volatility x 1.645 (variance-covariance method) |
| VaR backtesting exceptions | ≤ 5% of days (at 95% confidence) | Count(actual loss > VaR estimate) / total trading days |