What is Volatility?

By EC Assets Research Team · Published 19 August 2026 · Updated 19 August 2026

Volatility: The dispersion of an asset's returns, conventionally the annualised standard deviation of its log returns. It measures how far outcomes scatter, not which direction they take - which is why it is the input to option prices and position sizes, and a poor description of risk on its own.

What Volatility Actually Measures

Volatility is the dispersion of returns around their mean, stated as an annualised standard deviation. A 16 percent volatility means a one-standard-deviation annual move of roughly 16 percent in either direction.

The last three words carry the weight. Volatility is direction-blind: a stock that doubles and a stock that halves can carry identical volatility. It answers how widely outcomes scatter, never which way they fall. Treating it as a synonym for risk is the most common misuse of the number, and one an investor holding cash in a high-inflation regime can feel directly - no volatility, guaranteed loss.

Two versions of the number circulate, and they answer different questions. Realised volatility looks backwards and measures what the asset actually did. Implied volatility is extracted from option prices and states what the market is charging for future movement. Neither predicts the other reliably, and the gap between them is itself a tradable quantity.

How It Works

Volatility is computed from log returns, which are additive across periods and symmetric between gains and losses:

$$r_t = \ln\left(\frac{P_t}{P_{t-1}}\right)$$

The sample standard deviation of those returns, scaled to a year:

$$\sigma_{\text{annual}} = \sqrt{N} \cdot \sqrt{\frac{1}{n-1}\sum_{t=1}^{n}(r_t - \bar r)^2}$$

where $N$ is the number of periods in a year - 252 for US equity trading days. The square root comes from the scaling law: variance accumulates linearly with time, standard deviation with its square root.

The same asset, four horizons

A 16 percent annualised volatility, scaled by the square root of time. Empirical markets deliver 2σ periods considerably more often than the normal distribution allows.

Worked Example

A portfolio's daily log returns over a quarter have a standard deviation of 0.95 percent. Annualised:

$$\sigma = 0.95% \times \sqrt{252} \approx 15.1%$$

That single number then does a great deal of work. It sizes positions: at a 1 percent daily risk budget, the book is running roughly at its limit. It prices options: fed into Black-Scholes with a maturity and a strike, it produces a premium. It sets expectations: a 5 percent monthly drawdown is barely more than a one-sigma month and warrants no explanation, while the same move in a 5 percent volatility book is a three-sigma event that does.

When It Applies (and Limitations)

Volatility is not constant. It clusters - violent days follow violent days, quiet follows quiet. Mandelbrot observed this in 1963 and Engle formalised it in 1982; every modern volatility model is built on it. A single historical number describes an average of regimes that the asset may not currently be in.

It is not symmetric in practice. Equity volatility rises when prices fall and drifts lower when they rise, an asymmetry Black documented in 1976. Options price this in through the skew, which is why a single volatility figure never describes an equity option surface.

The normal distribution understates the tails. Volatility is a second moment; it says nothing about the third and fourth. Two assets with identical volatility can have entirely different crash profiles, which is why skewness, kurtosis and expected shortfall exist alongside it.

The estimate is noisy. Volatility measured over a short window carries substantial sampling error, and that error propagates into everything built on it - Sharpe ratios, value-at-risk, position sizes, volatility drag estimates.

Convention matters. Log versus simple returns, 252 versus 365 days, sample versus population standard deviation, close-to-close versus range-based estimators - each choice moves the answer by a few percent, and mixing conventions across a report is a silent source of disagreement.

Why It Matters for Institutional Investors

It is the common unit of risk. Position sizes, risk limits, capital allocations and stress scenarios are all expressed in volatility terms, which is what makes exposures in different asset classes comparable at all.

It prices optionality. Volatility is the one Black-Scholes input that cannot be observed, only estimated or implied. Every options desk is, at bottom, trading a view on this number.

It governs compounding. Because volatility drag scales with variance, two strategies with the same average return and different volatilities do not deliver the same wealth. Suppressing volatility raises the compounded outcome even when it leaves the average untouched.

It is the input to the wrong conclusion as often as the right one. A low-volatility asset is not a safe one: illiquid or infrequently marked holdings show low measured volatility precisely because their prices do not move, not because their risk is low.

References

1. Mandelbrot, B. (1963). The Variation of Certain Speculative Prices. Journal of Business, 36(4), 394–419.
2. Engle, R. F. (1982). Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation. Econometrica, 50(4), 987–1007.
3. Black, F. (1976). Studies of Stock Price Volatility Changes. Proceedings of the American Statistical Association.
4. Sinclair, E. (2013). Volatility Trading (2nd ed.). Wiley.