How do volatility models fit into distribution-based VaR estimation?

Question

AcadiFi · Accepted Answer

A volatility model describes how conditional volatility changes through time. A return distribution describes the standardized shock around that volatility. VaR typically needs both. A simple normal VaR may use forecast volatility and a normal quantile. A richer model may estimate conditional volatility plus a heavy-tailed or skewed innovation distribution.

Suppose Northfield Trading Risk fits a GARCH-style volatility model to daily equity returns. The model produces tomorrow's volatility forecast of 1.8 percent. If the analyst assumes normal standardized shocks, the VaR uses the normal quantile times that volatility. If the standardized residuals are fat-tailed, a t-distribution or other innovation model may produce a larger tail estimate.

The important distinction is conditional versus unconditional risk. Volatility clustering means tomorrow's risk depends on today's volatility state. Distribution fitting tells you how shocks behave after scaling by that volatility. FRM candidates should avoid thinking of volatility and distribution as substitutes. They are two pieces of one VaR model: one for time-varying scale, the other for tail shape. On the exam, also state the horizon, confidence level, data source, modeling assumption, and validation step, because those choices usually explain why two reasonable VaR answers differ materially. Practice more in our FRM question bank.

How do volatility models fit into distribution-based VaR estimation?

Master Part I with our FRM Course

Related Questions

Practice Questions