Downside Deviation

Standard deviation penalizes large gains and large losses equally. Downside deviation instead squares only the shortfalls beneath a reference return, typically zero, and takes the root of their mean, then annualizes it. It is the denominator of the Sortino ratio and is used when the concern is loss rather than movement generally.

How it is computed

sqrt(mean of squared min(daily return - minimum acceptable return, 0) over the trailing 90 days) x sqrt(365) x 100

Where it misleads

Only observations beneath the reference return enter the calculation, so a ninety-day window can leave well under half that many usable data points and the estimate becomes unstable in exactly the calm periods where it looks most reassuring. The minimum acceptable return is a provider choice, commonly zero but sometimes a cash rate or a target, and it is rarely disclosed even though it materially changes the result. Like standard deviation, the measure is blind to sequence: the same set of down days arranged as one abrupt collapse or as a slow grind yields the same number, though the consequences for a leveraged position differ completely. It also ignores gap risk from venue halts, where no down prints are recorded because trading stopped.

The equity comparison, and why it fails

Downside deviation as used in the Sortino ratio for funds; the estimator is identical, but equity and fund return series contain far fewer extreme observations, so the statistic is far better behaved there.

Reading a high or a low value

A higher figure means the days on which the asset declined were, on average, larger declines. Comparing it with total volatility shows how much of the movement came from the downside rather than the upside.