Options and volatility research

Variance risk premium, index-versus-constituent dispersion and dealer positioning: mechanisms that are well documented and expensive to implement honestly.

The variance risk premium is among the best documented effects in finance: implied volatility exceeds subsequent realised volatility on average, across markets and across decades, and there is a coherent economic reason why someone would pay it. Index-versus-constituent dispersion is its cross-sectional cousin, and dealer hedging flow is the microstructural residue of both. None of these is a secret.

They are, however, expensive to implement honestly, and that is where most of the published enthusiasm goes. Selling volatility is short convexity: the strategy earns a small premium repeatedly and gives it back in a single episode, so a backtest over any sample without such an episode is measuring the premium and not the risk. Add bid-ask spreads on options, the cost of delta hedging, and the assumption that a historical quote was executable, and a large fraction of the documented premium is gone before any of it is realised.

The feasibility question here is unusually binding. A point-in-time option surface includes open interest, quotes and the chain as it stood on a past date. It is either a licensed vendor product or a reconstruction, and a recent reconstruction cannot supply the years of observations a deflated result requires. That constraint is measured against the admission contract, so arithmetic decides before the idea does.

So what is published in this cluster is mostly boundary work: what the mechanism is, what it would take to test it properly, the execution contracts that would have to hold, and the crypto variance premium tested and killed. That is a thinner result than the literature would suggest, and it is the accurate one for a book that will not price a strategy at quotes it could not have traded.

The options and volatility documents