The Variance Risk Premium Is the Most Reliable Anomaly Nobody Talks About

Across decades, across indices, and across most liquid single names, implied volatility has traded above the volatility that subsequently materialised. Buyers of options have, on average, overpaid. The gap between what the option market charged for future variance and what future variance turned out to be is the variance risk premium, and it is among the most durable pricing regularities in finance.

The premium exists for a structural reason rather than a behavioural one. Volatility rises when portfolios fall, which means an instrument that pays off in high-volatility states is paying off precisely when the marginal dollar is most valuable to the holder. Investors will accept a negative expected return on such an instrument for the same reason they accept a negative expected return on insurance. The seller of that protection is compensated for accepting a payoff profile that is small and positive most of the time and large and negative occasionally.

Why the premium is not free money

The naive conclusion is that selling volatility systematically harvests a premium and therefore constitutes an edge. The arithmetic is correct and the conclusion is dangerous. The premium is compensation for bearing a specific risk, and that risk arrives in concentrated form. A strategy that collects a small premium in eighty-five percent of months and loses many multiples of the accumulated premium in the remaining fifteen percent has not found free money. It has found a payoff distribution whose average conceals its shape.

The practical distinction is between harvesting the premium with defined risk and harvesting it with undefined risk. Selling a put spread caps the loss at the width of the spread. Selling a naked put does not. Both collect a premium sourced from the same anomaly. Only one of them survives the event the premium exists to compensate for.

When the premium inverts

The variance risk premium turns negative during genuine dislocations, meaning realised volatility exceeds what implied volatility priced beforehand. This inversion is not a market failure, it is the moment the insurance pays out. Anyone modelling volatility strategies on a sample that excludes an inversion is modelling a business that has never encountered its own defining risk.

The inversions cluster. They arrive when positioning is crowded on the short side, which is to say after a long calm stretch during which the premium has been reliably harvested and capital has accumulated in the trade. This is the mechanism by which a strategy that looks safest is most fragile, and it is the reason positioning data belongs alongside the premium itself in any honest assessment of the opportunity.

Measuring the Premium Without a Terminal

A workable approximation compares the VIX at a point in time against the realised volatility of the index over the subsequent thirty calendar days, annualised. Repeating this across a long history produces a distribution of the premium rather than a single number, and the distribution is what matters: the mean is positive, the median is higher than the mean, and the left tail is far longer than the right.

That shape, a positive median with a long left tail, is the statistical signature of a short-insurance position. Any strategy whose backtest produces that shape should be understood as selling protection regardless of what instruments it uses or how its marketing describes it, because the payoff profile is the honest description of what a strategy does.