Value at Risk
Value at Risk (VaR) is a statistical risk measure that estimates the maximum potential loss of a portfolio over a given time horizon at a specified confidence level, under normal market conditions. A 1-day 99% VaR of $10 million means there is a 1% probability that the portfolio will lose more than $10 million on any given trading day.
Key takeaways
- VaR is defined by three parameters: the portfolio, the time horizon (typically 1 day or 10 days), and the confidence level (typically 95% or 99%).
- The three main VaR methodologies are parametric VaR (variance-covariance), historical simulation VaR, and Monte Carlo simulation VaR, each with different assumptions and computational trade-offs.
- VaR is used by banks under Basel II/III to calculate minimum regulatory capital requirements for market risk; the standardized VaR multiplied by a scaling factor sets the capital charge.
- VaR's most significant limitation is its inability to characterize tail losses beyond the confidence level—a 99% VaR tells you nothing about what happens in the remaining 1% of scenarios (the 'right-tail problem').
- Expected Shortfall (CVaR/ES)—the average loss in scenarios beyond the VaR threshold—has increasingly replaced VaR in regulatory capital frameworks (Basel III/FRTB) as it better characterizes tail risk.
Explanation
Value at Risk emerged in the late 1980s and early 1990s as a standardized risk communication tool at major financial institutions, most prominently at J.P. Morgan, whose 1994 publication of the RiskMetrics technical document codified the parametric (variance-covariance) VaR methodology and made it accessible to the broader industry. The appeal of VaR was immediate: it reduced complex, multi-dimensional portfolio risk into a single number—a maximum dollar loss at a given probability level—that could be communicated to boards, regulators, and investors without specialized quantitative knowledge.
The parametric VaR methodology assumes that portfolio returns are normally distributed and calculates VaR as: VaR = Portfolio Value × Z(α) × σ × √T, where Z(α) is the z-score corresponding to the confidence level (1.645 for 95%, 2.326 for 99%), σ is the daily portfolio volatility, and T is the time horizon in days. This formula is analytically tractable and computationally efficient—it only requires estimates of expected returns and the covariance matrix. However, the normality assumption is violated in practice: financial returns exhibit fat tails (higher probability of extreme events than the normal distribution implies), which means parametric VaR systematically underestimates tail risk, particularly during market crises when correlations spike and volatility clusters.
Historical simulation VaR avoids the normality assumption by applying the current portfolio's weights to historical return data—typically 250–500 trading days—generating a simulated distribution of portfolio returns. The VaR is then the percentile of this distribution corresponding to the chosen confidence level. Historical simulation naturally captures fat tails, volatility clustering, and correlation dynamics as they occurred historically, making it more robust for tail risk measurement. The limitation is that it only captures scenarios that actually occurred in the historical sample; rare but plausible scenarios not in the historical window—the 'black swan problem'—are excluded.
Monte Carlo simulation VaR generates thousands or millions of hypothetical scenarios by sampling from specified return distributions (which may be fat-tailed) and computing the portfolio value under each scenario. This approach is the most flexible—it can incorporate options, structured products with nonlinear payoffs, and complex correlation structures—but is computationally expensive and sensitive to model specification. For portfolios with significant nonlinear exposures (large options books, structured credit positions), Monte Carlo VaR is often the most appropriate methodology.
Despite its widespread adoption, VaR has been the subject of sustained methodological criticism. The 2008 financial crisis exposed its limitations dramatically: many bank VaR models estimated 99% 10-day VaRs of $100–$300 million while actual losses in individual trading days reached $1–$3 billion. The fundamental flaw is that VaR is coherent only under restrictive conditions and, crucially, is not sub-additive—the VaR of a combined portfolio can exceed the sum of the individual VaRs, violating the basic principle that diversification should reduce risk. Expected Shortfall (also called Conditional VaR or CVaR) is sub-additive and provides a complete characterization of tail risk by averaging all losses beyond the VaR threshold, making it a more robust risk measure. The Basel Committee's Fundamental Review of the Trading Book (FRTB) framework, finalized in 2019, replaced VaR with Expected Shortfall as the primary regulatory capital metric for banks' trading books.
Formula
Parametric VaR = P × Z(α) × σ × √T, where P = portfolio value, Z(α) = confidence level z-score, σ = daily volatility, T = time horizon; Historical VaR = percentile(α) of simulated P&L distribution
Example
A hedge fund's multi-asset portfolio has a current value of $500 million. Using historical simulation with 500 days of data, the fund's risk team ranks all 500 simulated daily P&L scenarios from worst to best. At the 99th percentile confidence level (top 1% of losses = bottom 5 scenarios out of 500), the 5th worst scenario shows a portfolio loss of $14.2 million. The fund's 1-day 99% VaR is therefore $14.2 million, or 2.84% of AUM. The Expected Shortfall at 99% (the average of the 5 worst scenarios) is $18.7 million—32% higher than VaR—reflecting the fat-tailed nature of the portfolio's loss distribution. The fund's risk committee sets a hard VaR limit of $15 million per day; the current portfolio at $14.2 million is close to this limit, prompting the risk manager to review the most VaR-contributing positions before adding new risk.
Related terms
Climate Risk Correlation Covariance Covariance Matrix Diversification Expected Shortfall Fat Tails Financial Crisis Hedge Fund Historical Simulation Var Monte Carlo Simulation Monte Carlo Var