Coefficient of Variation (CV)

Builds onSample Standard Deviation — if this page feels steep, start there.

CV = \frac{s}{\bar{X}}

$CV$: coefficient of variation — risk per unit of return, a pure number with no units
$s$: the standard deviation: how spread out the returns are
$\bar{X}$: the mean: the average return (the bar means 'average of')
$\frac{s}{\bar{X}}$: dividing risk by reward rescales the spread, so assets of any size can be compared

Why it must be true

Is 5% volatility a lot? It depends on how big the returns are. The coefficient of variation rescales dispersion by the mean, giving risk per unit of return — a pure number you can compare across assets with very different scales.

It is the mirror image of the Sharpe ratio: CV asks "how much risk per unit of reward?" (lower is better), Sharpe asks "how much reward per unit of risk?" (higher is better).

The derivation

Standard deviation and mean carry the same units, so their ratio is dimensionless:

CV = \frac{s}{\bar{X}}

Two consequences follow directly: CV is only meaningful when $\bar{X} > 0$ , and multiplying every observation by a constant leaves CV unchanged — which is exactly why it works across different scales.

When to reach for it

Ranking assets with very different scales of return by relative risk — risk per unit of reward, lower is better.

Listen for

“risk per unit of return”“relative variability”“compare dispersion across assets / funds”

Back-of-the-envelope

Estimate it in your head first — then the calculator only confirms.

≈
It's just 'risk per 1 of reward': σ 20% on mean 10% → CV 2. Round both to friendly numbers and divide.
≈
Cross-check direction: riskier-per-unit assets have BIGGER CVs. If the 'safe' asset came out higher, you inverted the fraction.

Traps in applying it

✗Inverting the ratio and ranking backwards.
✗Applying it when the mean is near zero or negative — the ratio explodes or loses meaning.
✗Comparing CVs built from different return frequencies.

Limits & criticisms

CV is undefined in spirit for negative means and unstable near zero. It makes no allowance for the risk-free rate — an asset earning barely above cash can look 'efficient' — which is exactly the gap the Sharpe ratio was built to close. It also inherits every limitation of standard deviation itself.

Where it came from

Karl Pearson introduced the CV in 1896 while comparing variability across populations with different scales — you cannot compare the dispersion of skull sizes and femur lengths in raw units, and you cannot compare the risk of a bond fund and a crypto fund in raw percentage points either.

Today CV survives as a quick screening tool — risk per unit of return, lower is better — and as the conceptual mirror of the Sharpe ratio, which flips the fraction and subtracts the risk-free rate.

One identity, 2 questions

The exam can hide any variable. Each face below is the same equation solved for a different unknown — drill them separately.

Risk per unit of return

CV = \frac{s}{\bar{X}}

The screening face: a pure number comparable across assets of any scale. Lower is better — the mirror image of Sharpe.

Drill this face →

Dispersion implied by a CV

s = CV \times \bar{X}

Read backwards: the CV is a multiplier on the mean, so the tolerable volatility scales directly with expected return.

Drill this face →

Drill this formula Back to the library