Geometric Mean vs Arithmetic Mean

Two averages, two different questions. The arithmetic mean adds its numbers and divides; the geometric mean multiplies them and takes the root. For most data the choice never arises — but the moment your quantities compound (growth rates, returns, ratios), only one of the two matches what actually happened to your money, your population, or your files.

The formula

geometric mean = ⁿ√(x₁ × x₂ × … × xₙ)
               = exp( (ln x₁ + … + ln xₙ) / n )

For 4 and 9: the product is 36 and √36 = 6 — exactly halfway between them on the multiplication scale, while the arithmetic mean 6.5 sits halfway on the addition scale. Both are legitimately "the middle"; they just measure different journeys. The logarithm form on the second line is what calculators actually evaluate, because it stays numerically stable even when the product of many terms would overflow into huge numbers. Every input must be strictly positive: logs of zero and negatives don't exist, which is why the geometric mean is undefined for data containing them.

The growth-rate mean: why −10% then +10% is not zero

The geometric mean earns its keep on percentages that compound. An investment loses 10% one year and gains 10% the next. The arithmetic mean of the returns is a reassuring 0% — but follow the money: 1,000 becomes 900, then 990. You lost 1%, and the "0% average" never told you.

factors: 0.9 × 1.1 = 0.99
geometric mean = √0.99 ≈ 0.995  →  −0.5% per year

The geometric mean of the growth factors, about 0.995, reports the honest average: −0.5% per period. This is volatility drag — any back-and-forth movement around a flat arithmetic mean quietly loses money, and the wider the swing, the bigger the loss. A fund advertising "averaged 8% a year" and a fund that actually multiplied your money at 8% compounded are different animals, and the gap between their arithmetic and geometric returns grows with volatility. For multi-year growth questions the same logic appears in finance as CAGR — the CAGR calculator computes it for start value, end value and years.

Equal distances, unequal in the other direction

Another intuition: on a multiplication scale, 4 and 9 are "equal distances" from 6, because 6/4 = 9/6 = 1.5. The geometric mean is the centre that makes ratios balance, exactly as the arithmetic mean makes differences balance. That is why it is the natural average for ratios of any kind — gear ratios, exchange rates applied in sequence, compression ratios, scale factors. Three enlargements of ×2, ×4 and ×8 average to ×4 (not ×4.67): 2 × 4 × 8 = 64 and ∛64 = 4, the factor that would get you the same end result in three equal steps.

The AM–GM inequality

For any set of positive numbers the two means stand in a fixed relationship:

geometric mean ≤ arithmetic mean
(equality only when every number is equal)

The gap between them measures spread from a multiplicative point of view — identical numbers give a ratio of exactly 1, while widely varying numbers pull the two apart. That makes the ratio a useful diagnostic in its own right: if you compute both means for a return series and find a large gap, volatility is eating your average, and no change of arithmetic will recover it. It also gives an elegant proof tool: many maximum-product problems ("which rectangle of perimeter 20 encloses the most area?") reduce to AM ≥ GM, with the answer "when all factors are equal."

When to use which mean

Choose by how the quantities interact. Quantities that add — salaries, heights, commute times — want the arithmetic mean. Quantities that multiply — successive growth rates, ratios, index levels, discount factors — want the geometric mean. (A third sibling, the harmonic mean, governs rates over fixed quantities, like average speed over equal distances.) And when the data is skewed or full of outliers, remember that both means are hostage to extremes; the median is the resistant alternative, as covered in the mean, median and mode guide. For the general-purpose workhorse, see the Average Calculator.

Try it yourself

Paste your numbers into the Geometric Mean Calculator — it computes the geometric mean via the log formula, shows the product, and displays the arithmetic mean and the AM/GM ratio beside it so you can see the compounding gap at a glance. Try 4 and 9 for the clean √36 = 6, then try 1.1 and 0.9 to watch the volatility-drag effect appear in a single line.