20% off Lifetime $320 instead of $400 with code . Ends 24 Nov (2d 23h)

See Lifetime

WSU Lexicon: Quantitative Methods

Target Downside Deviation

Target downside deviation is a risk measure that captures only the dispersion of returns falling below a chosen target, such as zero or a minimum acceptable return. It squares each shortfall below the target, sums them, divides by the sample size minus one, and takes the square root.

When the target equals the sample mean, the measure is called semideviation, and its square is semivariance.

The entry

Computing the Measure from a Return Sample

Start by fixing the target, labelled B. For each period, compare the return with B. Where the return is below B, square the difference between the return and B; where it is at or above B, record zero. Add the squared shortfalls, divide that sum by the total number of observations minus one, and take the square root to return to percentage units.

Consider five annual returns of 8%, negative 4%, 2%, negative 6% and 10% with a target of 0%. Only the two losses fall short, giving squared shortfalls of 16 and 36, a total of 52. Dividing by four, the sample size less one, produces 13, and the square root gives a target downside deviation of about 3.61%. The ordinary standard deviation of the same series is about 7.07%, roughly twice as large, because the 8% and 10% gains also contribute to measured dispersion.

The entry

Why Separate Downside from Upside Dispersion

Standard deviation treats a return five percentage points above the mean exactly like one five points below it. For an investor whose concern is loss or failure to meet a liability, that symmetry is unhelpful: upside surprises are welcome, not risky. Target downside deviation isolates the outcomes that actually hurt.

The distinction matters most when return distributions are asymmetric. Strategies that sell options, carry credit exposure or trade merger spreads tend to produce frequent small gains and occasional large losses, a negative skew that standard deviation can understate as a description of pain. Conversely, a long-option or trend-following profile may show high standard deviation driven mainly by gains. For a symmetric distribution, semivariance is roughly half the variance, so the downside measure adds little; its value lies in skewed data.

The entry

Choosing the Target and Reading the Result

The target is a modelling choice, and the number is meaningless without it. Common choices are zero, which frames risk as any loss; the risk-free rate; a minimum acceptable return drawn from an investment policy statement, such as a pension plan's required return; and the sample mean, which yields semideviation. Raising the target pulls more observations into the shortfall group and enlarges each shortfall, so the measure rises with the target.

Comparisons across funds are valid only when the same target and the same measurement period are used. Because the result is expressed in return units, it can be read alongside the mean: a fund averaging 6% with a downside deviation of 4% relative to zero has less typical shortfall than one with the same mean and a downside deviation of 9%.

The entry

What Candidates Must Be Able to Do

Expect to compute the measure from a short return series given a target, which means filtering correctly, keeping the full sample size in the denominator and remembering the final square root. Candidates should also recognise the related vocabulary: semivariance is the squared version using the mean as target, semideviation is its square root, and target semivariance is the squared version of target downside deviation.

Interpretive questions typically ask which of two portfolios carries more downside risk, or why a manager with negatively skewed returns looks safer on standard deviation than on downside deviation. The measure also feeds the Sortino ratio, computed as mean return minus the target, divided by target downside deviation, which candidates should be able to contrast with the Sharpe ratio.

Common trap

“The sum of squared shortfalls is divided by the number of returns below the target, not by the full sample size.”

In the convention candidates are tested on, the denominator is the total number of observations minus one. Returns at or above the target still belong to the sample; they simply contribute zero to the numerator. Dividing by the below-target count alone overstates downside risk.

In practice

The Sortino ratio's denominator

Frank Sortino popularised a variant of the Sharpe ratio that divides the excess of mean return over a minimum acceptable return by target downside deviation instead of standard deviation. The resulting Sortino ratio is a standard statistic in fund analytics and hedge fund reporting.

Candidates ask

Questions on this term

  • How does target downside deviation differ from standard deviation?Standard deviation measures dispersion on both sides of the mean. Target downside deviation counts only returns below a chosen target, so above-target outcomes add nothing to the measured risk.
  • Is semideviation the same as target downside deviation?Semideviation is a special case. It uses the sample mean as the threshold, so it measures the dispersion of below-average returns. Target downside deviation allows any threshold, such as zero, the risk-free rate or a minimum acceptable return. When the target is set equal to the mean, the two calculations produce identical results.
  • What happens to returns that exactly equal the target?They contribute nothing to the numerator, because their shortfall is zero, but they remain part of the sample and are therefore counted in the n minus one denominator. The same is true of every return above the target. Only the numerator is filtered; the denominator always reflects the full number of observations in the sample.

See also

Related entries

  • Sharpe RatioThe Sharpe ratio measures excess return per unit of total risk: the portfolio's return minus the risk-free rate, divided by the portfolio's standard deviation.
  • SkewnessSkewness measures the asymmetry of a probability or return distribution around its mean.
  • Standard DeviationStandard deviation is the square root of variance: a measure of how widely observations disperse around their mean, expressed in the same units as the data.
  • VarianceVariance is the expected squared deviation of a variable's outcomes from its mean, the standard measure of dispersion in finance.

One entry from the WSU Lexicon every day. Browse the full archive.