VEGA FIELD GUIDE · 08 · VOLATILITY
The variance risk premium
Options often price more variance than the underlying later delivers. That cushion pays for taking convexity and jump risk off someone else’s hands. Its size, history, and event context matter far more than its mere existence.
Variance risk premium is most naturally expressed as implied variance minus expected or realized variance.
A positive premium is common and can widen during stressed or event-heavy markets.
Level, percentile, and z-score answer different questions about richness.
Why option variance can carry a premium
Investors buy options for convex protection, event exposure, and balance-sheet certainty. Sellers accept nonlinear losses and the practical burden of hedging. The price difference between option-implied variance and subsequently delivered variance is commonly called the variance risk premium, or VRP.
The premium changes with fear, positioning, liquidity, macro uncertainty, and company events. It can remain positive for long stretches and then be overwhelmed by a single jump. That distribution is exactly why the premium exists.
Define the comparison before reading it
A backward-looking version compares today’s implied variance with recent realized variance. A forward evaluation compares entry implied variance with variance realized over the option’s life. A forecast version compares implied variance with a model estimate for the matching future horizon.
Vega labels the components and lookbacks. For a current screen, IV30 versus HV20 is a practical regime proxy. Historical trade economics and delta-hedged replay provide a closer view of whether a standardized position captured the apparent edge.
VRP = IV_horizon^2 - expected realized volatility_horizon^2Level, relative rank, and z-score
The raw variance difference shows economic magnitude. A historical percentile shows how often the premium has been smaller. A z-score measures distance from the trailing mean in units of its own standard deviation. Cross-sectional rank shows where the name sits among peers on the same date.
Each lens adds something. A high z-score based on a tiny absolute premium may have limited trade value after spreads. A large raw premium with an ordinary z-score may be normal for a company that regularly crosses earnings and product events.
| Lens | Question | Useful follow-up |
|---|---|---|
| Raw VRP | How much variance is priced? | Compare with spread and event cost |
| Percentile | How unusual is this through time? | Inspect the historical distribution |
| Z-score | How far from the usual regime? | Check outliers and sample size |
| Cross rank | How rich versus peers today? | Filter sector and liquidity |
Worked example: price the cushion
Suppose 30-day ATM IV is 28% and a matching realized-volatility estimate is 22%. Implied variance is 0.28² = 0.0784. Expected realized variance is 0.22² = 0.0484. The annualized premium is 0.0300.
Over 30 days, the simple time-scaled variance difference is 0.0300 x 30 / 365 = 0.00247. IV is 27.3% above the volatility estimate on a ratio basis, yet that percentage is only a headline. The option’s pathwise P/L also responds to entry skew, gamma, theta, vega, jumps, transaction costs, and hedge timing.
0.28^2 - 0.22^2 = 0.0300 annualized varianceFrom premium to a candidate
- Use Pulse to see VRP breadth and the richest or cheapest current extremes.
- Screen by VRP level or z-score, then filter for liquidity, sector, market-cap tier, and earnings distance.
- Open the symbol to inspect term structure and skew. The premium may live in one tenor or wing.
- Choose a structure whose Greeks match the thesis. Defined-risk spreads, condors, and calendars carry different exposures.
- Check calculator economics and historical replay after bid-ask spread and fees. Premium capture lives in net results.
Short-volatility returns can look smooth until a gap arrives. Size the position around its loss distribution and maximum risk.
See variance-premium breadth and extremes
Scan the market, open a candidate, and carry the research into a defined calculator position.