Coefficient of Variation

The calculation is a single division: take the standard deviation of the data series and divide it by the arithmetic mean of that same series. Both inputs must be measured in identical units and over identical periods, so the units cancel and the result is a pure number, frequently reported as a percentage by multiplying by 100. If returns are measured monthly, the standard deviation must be monthly as well; mixing an annualised volatility with a monthly mean produces a figure with no interpretation.

At Level I, candidates work with sample data, so the sample standard deviation, using n minus one in the denominator of the variance, pairs with the sample mean. The reciprocal of the coefficient of variation is sometimes quoted as a signal-to-noise ratio: mean per unit of dispersion, where higher is better. The two forms rank alternatives identically, but in opposite directions, and confusing the direction of preference is a common source of error.

Portfolio managers use the measure to screen candidate funds, strategies or asset classes whose return levels differ by an order of magnitude, since raw volatility unfairly penalises high-return, high-volatility assets. Equity analysts apply it to fundamental series: the variability of a company's quarterly revenue, earnings or operating margin relative to its average level, which gives a scale-free read on earnings quality across firms of very different size.

Outside investment selection, the statistic is standard in cost and quality control, where the variability of unit costs, defect rates or delivery times is compared across production lines with different baselines. Credit analysts use it on cash-flow series to compare the stability of borrowers of unequal size. In every case the appeal is the same: the denominator normalises for scale so that two series measured on different magnitudes become directly comparable.

The Sharpe ratio divides mean return in excess of the risk-free rate by the standard deviation of returns. The coefficient of variation divides standard deviation by mean return without subtracting anything. So the two are close relatives: ignoring the risk-free rate, the Sharpe ratio is approximately the reciprocal of the coefficient of variation. The difference matters. When the risk-free rate is meaningful, the Sharpe ratio measures compensation for bearing risk, while the coefficient of variation simply measures relative dispersion around whatever average the series happens to have.

Both statistics rely on standard deviation, so both treat upside and downside deviations symmetrically and both are distorted by fat tails and skew. Neither says anything about the shape of the distribution, and neither is reliable on short samples where the estimated mean is unstable.

Expect to be given two or more series, each with a mean and standard deviation, and asked which has greater relative risk or dispersion. Compute the ratio for each, then select the smaller value as the less disperse alternative. Questions frequently supply raw observations instead, requiring the sample mean and sample standard deviation to be calculated first, so the n minus one convention should be automatic.

Be prepared for the conceptual variant as well: identifying when the measure is appropriate. It requires a ratio-scale variable with a meaningful, positive mean. Negative or near-zero mean returns, which occur regularly in real return series, make the ratio unstable or meaningless, and the correct answer in such items is that the comparison should not be made on this basis. Stating the interpretation cleanly, as risk carried per unit of average return, is often enough to eliminate distractors.