Volatility is the standard measure of risk in finance. It is used to size positions, construct portfolios, set capital requirements and price options. It is also a measure of how much prices have moved recently, which is a different thing from how much you might lose — and the gap between the two opens widest precisely when it matters most.
What volatility measures
Volatility is the standard deviation of returns: the dispersion of price changes around their average, usually annualised.
It became the default risk measure for practical reasons. It is calculable from price history alone, it aggregates across a portfolio through well-understood mathematics, and it is symmetric and tractable. Nothing about it was chosen because it captures what investors actually fear.
The mismatch is straightforward. An investor does not fear that prices will move. They fear a permanent loss of capital. Those coincide often enough for volatility to be a serviceable proxy in normal conditions, and they come apart in exactly the situations that matter.
Three specific failures
It treats upside and downside identically. A position that rises 20 percent contributes as much measured volatility as one that falls 20 percent. Nobody experiences those symmetrically. Measures such as downside deviation and maximum drawdown address this, and are less used because they are less mathematically convenient.
It is backward-looking. Calculated from historical prices, it describes the period it was measured over. Risk concerns the future. Volatility persists in the short run — calm periods tend to be followed by calm periods — which makes it a reasonable near-term estimate and a poor guide to regime change. It gives no warning of the transition, and the transition is the event.
It cannot see what has not happened. An asset that has been stable for three years has low measured volatility whether it is genuinely safe or has simply not yet encountered the event that will reprice it. A currency peg is perfectly stable until it breaks. A credit instrument that has never defaulted has no default history. Volatility measured over a period containing no shock describes a period containing no shock.
Why low volatility can create risk
The most important failure is not that volatility understates risk during calm periods. It is that low volatility can cause risk to build.
A substantial amount of capital is managed to a volatility target. When measured volatility falls, these strategies increase leverage — mechanically, to maintain the target. Risk models used for capital allocation behave the same way: lower measured volatility permits larger positions on the same capital.
So a prolonged calm period produces higher leverage across the system. When volatility eventually rises, the same mechanism reverses: positions must be reduced, and everyone reduces simultaneously because they are all responding to the same observation. Forced selling raises volatility further, which requires further selling.
The instability is endogenous. The measure is not merely failing to detect the risk — its widespread use as a control variable is part of what generates it. This is the mechanism behind the observation that stability is destabilising, and it is the main reason to be uneasy about extended periods of unusually low volatility.
The correlation problem
Portfolio construction depends on correlations as much as volatilities. Diversification works because assets do not move together, so a portfolio’s volatility is lower than the average of its holdings.
Correlations are estimated from history, and they are not stable. Specifically, they tend toward one during severe market stress — the assets that diversified each other for years fall together in the week that diversification is needed.
The reason is that in a liquidity event, the question stops being what an asset is worth and becomes what can be sold. Investors sell what they can rather than what they want to, which transmits selling into assets with no fundamental connection to the original problem.
A portfolio’s measured risk therefore understates its true risk by more than any single position’s volatility suggests, and the understatement is largest during crises.
What to use instead
Not nothing — volatility is genuinely useful, and the alternative is not abandoning measurement.
- Maximum drawdown. The largest peak-to-trough decline actually experienced. It corresponds directly to what an investor feels and to whether they will abandon a strategy at the worst moment.
- Leverage. Explicit and embedded. Leverage determines whether a drawdown is a bad period or a terminal event, and it is knowable in advance rather than estimated.
- Liquidity of the holdings. How long it would take to exit at a reasonable price. This is not captured by volatility at all, and illiquid assets frequently show low measured volatility precisely because they are not repriced often.
- Scenario analysis. What specifically would have to happen for a permanent loss, and how plausible is it. Slower than a single number and considerably more informative.
The general discipline is to distrust any single number that compresses risk into one figure, and to be most suspicious when that figure is reassuringly low. A quiet market is not the same as a safe one — and the mechanism that makes it quiet is often the same one that will make it disorderly.
