A statistical risk measure that, for a specified confidence level α and time horizon, identifies the loss threshold L such that the probability that portfolio losses over the horizon exceed L equals 1−α; operationally, VaRα is the α‑quantile of the portfolio loss distribution for that horizon and model assumptions.

Principle

VaR summarizes downside risk as a loss distribution quantile for a chosen horizon and confidence; it does not quantify the magnitude or probability distribution of losses beyond that quantile.

Demonstration

Illustrative scenario → A trader computes a one‑day 95% VaR of $1,000,000 for a portfolio under a specified model. Interpretation: given the model and horizon, there is an estimated 5% chance of a loss exceeding $1,000,000 on a single day; VaR does not state how large losses in that 5% tail might be.

Misapplication

Mistaken interpretation → Treating VaR as the worst possible loss, assuming subadditivity (ignoring that VaR may not be coherent for some loss distributions), or relying on a single VaR figure without accounting for model risk, parameter uncertainty, or tail behavior. Error: confusing a quantile summary with a complete description of tail risk.

Consequence

VaR provides a concise risk limit or reporting metric that can guide risk appetite, capital allocation, and control. Its use can also create incentives to shift exposure toward low‑probability, high‑impact tail events if tail severity is ignored; effective risk management supplements VaR with stress tests and tail‑sensitive measures.

Reversal

Qualification → For heavy‑tailed or non‑stationary loss processes, VaR can be unstable or misleading; tail‑expectation measures (e.g., expected shortfall) and scenario analysis are more informative about extreme losses and diversification effects.

Boundary

Scope → Clearly within: quantile‑based assessment of portfolio loss for a defined horizon, confidence level and model. Boundary case: non‑linear portfolios with options where horizon aggregation and model choice materially affect VaR. Clearly outside: full tail‑risk descriptions such as expected shortfall or distributional tail fitting.

Semantic Tension

Tension → Simplicity and communicability of a single quantile summary versus the need for sensitivity to tail severity, model uncertainty and diversification properties.

Synthesis

VaR is a compact, model‑dependent quantile that communicates downside probability over a horizon; it is useful as a summary constraint but must be interpreted alongside tail metrics, stress scenarios, and an explicit account of model assumptions and uncertainties.