Why selling volatility is not itself an edge

Short-volatility strategies have a feature that makes them unusually easy to misread: on average, they are paid for doing nothing clever. This note separates the premium that selling volatility earns from the contribution a manager might add to it, and describes how that contribution can be measured.

The premium is real

On equity indices, implied volatility has on average exceeded the volatility subsequently realized. The difference, the variance risk premium, is among the best-documented regularities in derivatives markets (Bakshi and Kapadia 2003; Carr and Wu 2009). Its mirror image appears in option returns: positions that are long volatility have tended to lose money on average, even when their market exposure is hedged (Coval and Shumway 2001).

The explanation is not an inefficiency. Volatility tends to rise when markets fall and wealth is scarce. Investors will pay to hold assets that pay off in those states, and a short-volatility position is the other side of that trade: it earns steady income in ordinary conditions and takes its losses when conditions are worst. The premium is compensation for the timing of those losses.

That explanation has two consequences. Because the premium compensates a real risk, it should persist. And because it compensates a real risk, it should, from time to time, hurt badly. In February 2018 a single-session doubling of the VIX index led to the termination of one of the largest inverse-volatility exchange-traded notes. That was not the premium failing. It was the premium's other side.

Why the premium is not an edge

Anyone can hold the exposure. Index put-writing benchmarks, short positions in volatility futures and short-volatility exchange-traded products all harvest some version of the same premium. The return accrues to the exposure, not to the person holding it.

A short-volatility track record therefore mostly measures two things: the premium, and the particular sequence of states the sample happened to contain. Samples without a severe volatility shock overstate what the exposure earns, because the losses that justify the premium are missing from them. A long record with good statistics is compatible with no skill at all.

Where a manager can add something

What a manager can contribute is the decision about when, and how much, volatility risk to hold.

That decision has economic content because the premium itself depends on the state. The volatility futures curve usually slopes upward, and a short position then earns roll yield as futures converge toward the lower spot level. When the curve inverts, usually in stress, the roll works against the short position at the moment the risk of further spikes is greatest. A policy that sizes exposure by state is making a claim about how compensation and risk vary across states. That claim can be right or wrong, and it can be tested.

How to measure the contribution

The right benchmark for a state-dependent short-volatility strategy is not cash or an equity index. It is the same exposure held without the policy.

In practice that means at least two baselines:

1. Unconditional exposure. The same instrument held continuously, at a size matched to the strategy's average exposure.
2. A naive state rule. A simple, pre-specified rule of the kind any practitioner would try first, such as reducing exposure when the futures curve inverts.

The manager's contribution is the difference between the policy and these baselines, evaluated out of sample, with particular attention to the states in which losses occur. Averages can mislead here. Two policies with similar average returns can differ sharply in their exposure on the handful of days that decide the outcome.

This is also where research discipline matters more than model sophistication. The states that matter most are the rarest. Any sample contains only a few major volatility shocks, and a policy that has been tuned, even informally, on those episodes has been tested on the data that selected it. The effective sample for the question that matters most is a few independent episodes. Pre-specifying the rule, recording what was tried, and holding out evidence that has not been looked at are what make a claim of added value credible.

What a description should say

A description of a short-volatility strategy should state:

• the premium it harvests, and why that premium should exist;
• the states in which the premium is earned and the states in which losses occur;
• the policy, and how exposure changes across those states;
• the benchmark exposure against which the policy is judged;
• the evidence that the policy adds something the unconditional exposure does not, labeled by kind: simulated, out of sample or observed in production; and
• its failure states, including moves that arrive faster than the policy can respond, liquidity in the instruments used, and the structural risks of the products themselves.

Selling volatility is a way to be paid for bearing a specific, well-understood risk. It becomes a strategy, and possibly an edge, only through the policy that decides when and how much of that risk to bear. That contribution has to be measured against the exposure it is built on.

References

Bakshi, G., and N. Kapadia (2003). "Delta-Hedged Gains and the Negative Market Volatility Risk Premium." Review of Financial Studies 16 (2): 527–566.

Carr, P., and L. Wu (2009). "Variance Risk Premiums." Review of Financial Studies 22 (3): 1311–1341.

Coval, J. D., and T. Shumway (2001). "Expected Option Returns." Journal of Finance 56 (3): 983–1009.