Weighted Mean and Geometric Mean: Beyond the Simple Average

The ordinary mean treats every value as equally important — but that's not always true. A course grade weighted toward the final exam, or a growth rate that compounds over several years, both need a different kind of average. The weighted mean and geometric mean are the two most common alternatives to the arithmetic mean, each built for a specific situation the plain average handles poorly.

The Weighted Mean: Not Every Value Counts Equally

These extend the same core idea as mean, median, and mode — different answers to "what's typical" — into situations the plain arithmetic mean handles poorly: unequal importance, and multiplicative compounding.

Each value gets multiplied by its own weight before averaging, then divided by the total weight rather than a simple count. Worked example: a course grade is built from a midterm (worth 30%) and a final exam (worth 70%), scored 70 and 90 respectively. x̄w = (0.30×70 + 0.70×90) / (0.30+0.70) = (21+63)/1 = 84 — noticeably different from the unweighted average of 80, correctly reflecting that the final exam counted for much more of the grade.

Weighted means show up constantly once you know to look for them: a stock portfolio's average return weighted by how much money is invested in each holding; a survey's results weighted to correct for over- or under-sampling certain demographic groups; a student's GPA weighted by each course's credit hours. Using an unweighted mean in any of these cases silently misrepresents the true picture.

The Geometric Mean: Built for Rates and Ratios

Instead of summing values and dividing by n, the geometric mean multiplies them together and takes the nth root. This is the correct averaging method whenever the values being combined are themselves rates, ratios, or growth multipliers — situations where values compound rather than simply add together.

Worked Example: Why the Arithmetic Mean Gives the Wrong Answer for Growth

Problem: An investment grows 50% in year one, then falls 50% in year two. What's the average annual growth rate?

Arithmetic mean (incorrect for this purpose): (50% + (−50%))/2 = 0% — suggesting the investment ended up unchanged.

What actually happened: starting with $100, a 50% gain brings it to $150; a 50% loss then brings it down to $75 — the investment actually LOST 25% overall, not stayed flat, despite the arithmetic mean of the two percentage changes being exactly 0%.

Geometric mean (correct): convert each year's change to a growth multiplier — 1.50 and 0.50 — then GM = √(1.50 × 0.50) = √0.75 ≈ 0.866, meaning an average annual multiplier of 0.866, or roughly −13.4% per year. Compounding −13.4% twice: 0.866 × 0.866 ≈ 0.75, correctly landing on the true 25% overall loss. The geometric mean correctly captures the compounding nature of sequential growth rates in a way the arithmetic mean structurally cannot.

When to Use Each Type of Mean

• Arithmetic mean (covered in mean, median, and mode) — the default for genuinely additive quantities where every observation counts equally: test scores, heights, most everyday measurements.
• Weighted mean — when observations genuinely deserve different importance: grades with different component weights, survey data needing demographic correction.
• Geometric mean — rates of change, growth rates, ratios, or any quantity that compounds multiplicatively across periods rather than adding.

Real-World Uses

Finance: compound annual growth rate (CAGR), a geometric mean application, is the standard way to report investment performance over multiple periods, precisely because it correctly reflects compounding.

Economics: inflation rates and GDP growth rates, when averaged across multiple years, use the geometric mean for the same compounding reason.

Academic grading: weighted means are standard practice for combining assignments, exams, and participation into a single course grade, with weights set by the syllabus.

Survey research: weighted means correct for known sampling imbalances, ensuring a survey's overall results properly reflect the true population composition even when certain groups were over- or under-represented in the raw sample.

Common Mistakes

Averaging percentage changes with the arithmetic mean. As the worked example shows, this is a genuinely common and consequential error — not just in finance, but anywhere sequential percentage changes are being summarized (population growth rates, price changes over time, success rates across repeated processes).

Forgetting to normalize weights. The weighted mean formula divides by the SUM of weights, not by the count of observations — using weights that don't sum to a meaningful total (or forgetting the division entirely) produces a result on the wrong scale.

Using a weighted mean without justifying the weights. Weights should reflect a genuine, defensible reason one observation matters more than another (like exam weighting stated in a syllabus, or population proportions in survey correction) — choosing weights arbitrarily, or worse, choosing them specifically to produce a desired result, undermines the legitimacy of the calculation.

A Second Worked Example: Weighted Mean in Portfolio Returns

Problem: An investment portfolio holds $6,000 in Stock A (return: 8%) and $4,000 in Stock B (return: 12%). What's the portfolio's overall weighted return?

Weights: $6,000/$10,000 = 0.6 for Stock A; $4,000/$10,000 = 0.4 for Stock B.

Weighted return = (0.6 × 8%) + (0.4 × 12%) = 4.8% + 4.8% = 9.6%

This is meaningfully different from a naive unweighted average of the two returns (10%), correctly reflecting that more money was invested in the lower-returning Stock A. A weighted mean is essential here — reporting a simple average of the two percentage returns would misrepresent the portfolio's actual overall performance.

A Third Mean Worth Knowing: The Harmonic Mean

Beyond the arithmetic, weighted, and geometric means, a fourth type — the harmonic mean — is the correct averaging method for rates expressed as "quantity per unit," like speed (distance per time) when averaging across equal distances rather than equal times. If you drive a fixed distance at 60 mph one way and 40 mph on the return, the average speed for the full round trip is the harmonic mean of 60 and 40 (48 mph), not their arithmetic mean (50 mph) — because you spend more TIME at the slower speed, which the harmonic mean correctly accounts for and the arithmetic mean does not.