Can you average percentages? Yes — sometimes. A percentage is not a number like any other; it is a ratio with its base hidden, and averaging percentages without asking about those bases is one of the most common arithmetic errors in school, work and news reports. This guide gives the two honest methods: when a plain average of percentages is exactly right, when it is quietly wrong, and the weighted fix that works in both worlds.

How to calculate average percentage when the bases are equal

Start where the plain method is correct. Three quizzes, each marked out of the same 20 questions, come back as 16, 18 and 15 correct — that is 80%, 90% and 75%. Their average percentage is (80 + 90 + 75) ÷ 3 = 245 ÷ 3 ≈ 81.67%. Now verify it the way every average of percentages deserves to be verified — by pooling the raw counts: total correct 16 + 18 + 15 = 49, total questions 20 + 20 + 20 = 60, and 49 ÷ 60 ≈ 81.67%. The two methods agree to the last digit. They agree because each percentage was built on the same base: every point of percentage represented exactly 0.2 of a question in each quiz, so the percentages were already weighted equally before anyone averaged them.

That is the general principle to hold onto: averaging percentages directly is legitimate exactly when the bases behind them are equal — same number of questions, same amount invested, same-sized groups. Test scores on identical papers, monthly defect rates on identical production runs, shooting percentages over identical attempts: in all of these, how to find average percentage is just “add them and divide by the count”, and the pooled check will always confirm it.

How to average percentages when the bases are not equal

Now break the condition and watch the plain method fail. A student scores 90% on a 10-question quiz and 50% on a 200-question examination. The naive average of percentages is (90 + 50) ÷ 2 = 70%. It looks reasonable and it is badly wrong. Rebuild what the percentages stand for: 90% of 10 is 9 correct; 50% of 200 is 100 correct. Pooled honestly, the student answered 9 + 100 = 109 of 10 + 200 = 210 questions, which is 109 ÷ 210 ≈ 51.9%. The examination outweighs the quiz twenty to one in reality; the naive average pretended they were equals, and the pretence was worth 18 points.

This is the shape of nearly every real-world percentage error. Hospital success rates pooled across a small clinic and a large hospital, conversion rates across a quiet week and a launch week, a fund’s yearly returns across years when different amounts were invested — in each case, percentages of unequal bases were averaged as if the bases matched. The percentages themselves were never wrong. The averaging was.

The weighted fix, step by step

The repair is mechanical, and it is the same weighted mean used for grades and credit hours. Here is how to find the average of percentages with unequal bases:

1. Recover each base and each part. A percentage plus its base gives the part: 90% of 10 questions = 9 correct.
2. Add all the parts into one total part (9 + 100 = 109).
3. Add all the bases into one total base (10 + 200 = 210).
4. Divide once, at the end: 109 ÷ 210 ≈ 51.9%.

One division, performed last, on real counts. That is the whole fix — and notice it degrades gracefully: when the bases happen to be equal, steps 1–4 return exactly the plain average, which is why the pooled check in the first section agreed.

A money example, because percentages of money are where this bites adults. In January you save $200 of a $1,000 income — a 20% savings rate. In February you save $450 of a $1,500 income — 30%. The naive average says 25%. The honest two-month rate pools the dollars: ($200 + $450) ÷ ($1,000 + $1,500) = $650 ÷ $2,500 = 26%. One point of difference — the reward for remembering that February’s percentage was standing on a base half again as large. This is also how to find the average percentage for a whole year of rates, returns or discounts: never average the printed percentages; pool what they were percentages of.

Average of two percentages: the simplest case

Two percentages tempt people most, because the arithmetic looks trivial. If both stand on the same base — two halves of one match, two sections of one exam — the average of two percentages is just their midpoint: 40% and 60% average to 50% — no apology needed. But the quiz-and-exam contrast above was also a two-percentage problem, and its honest answer (51.9%) sat nowhere near the midpoint of 90 and 50. The number of percentages is never the question; the bases always are. A useful reflex: whenever someone hands you an average of exactly two percentages, ask “of what?” twice — once for each.

Three neighbours of this mistake worth knowing

Averaging percentages has close cousins that catch the same people. Stacked discounts: 20% off followed by 10% off is not 30% off. On a $100 item the first discount leaves $80, and the second takes 10% of $80, leaving $72 — a total saving of 28%, because the second percentage stood on a smaller base. Successive percentages multiply; they never add.

Percentage points versus percent: when a rate itself moves from 4% to 5%, it has risen by 1 percentage point — and by 25% relative to its old value ((5 − 4) ÷ 4 = 0.25). Both statements are true; they measure different things. Reports that swap them are either careless or hoping you will not notice, and the base — the old rate — is once again the hidden quantity that decides which is meant.

Growth rates: an investment that gains 50% one year and loses 50% the next has not broken even, whatever the naive average of +50 and −50 suggests. $100 grows to $150, then falls to $75 — a 25% loss overall. Percentage changes compound on a moving base, so their honest average is a multiplicative one (the geometric mean), a reminder that “average the percentages” fails in a second, distinct way whenever the percentages are changes rather than shares.

Checking someone else’s average percentage

The pooled method is also an audit tool. Suppose a school reports that “the average pass rate across our two final-year classes is 75%”. The classes: 30 students at a 90% pass rate and 10 students at a 60% pass rate. The reported 75% is the naive midpoint of 90 and 60 — but pooled honestly, passes are 0.90 × 30 = 27 and 0.60 × 10 = 6, so the true combined rate is (27 + 6) ÷ (30 + 10) = 33 ÷ 40 = 82.5%. The school’s figure understated its own results by 7.5 points, because the bigger class was also the stronger one. Run the same check on any headline average of percentages — class results, branch sales growth, hospital rates — and you will find the naive midpoint doing quiet damage in a surprising share of them. The tell is always the same: an average of percentages quoted without the bases that produced them. Ask for the bases, pool the parts, divide once. If the bases are refused or unavailable, treat the quoted average as decoration rather than data.

One more equal-bases case: average discount on a basket

To end on the friendly side of the rule, here is an everyday case where the plain average is exactly right. Two items, each priced at $50: one is discounted 10%, the other 30%. The savings are $5 and $15, totalling $20 on a $100 basket — an average discount of 20%, which is precisely the plain average of 10 and 30. The bases were equal ($50 each), so the percentages deserved equal votes, and pooling confirms the shortcut. Change either price and the equality collapses — a $20 item at 30% off beside a $180 item at 10% off saves $6 + $18 = $24 on $200, an average discount of just 12%, nowhere near the naive 20%. Same percentages, different bases, different truth: the refrain of this entire guide, in a shopping basket.

The decision table

The third row is the one nobody wants to print: sometimes the correct answer to “what is the average percentage?” is that the information given cannot produce one.

Whichever name a searcher uses for this — percentage average calculator, average calculator percentage, percentage calculator average — the tool they need is doing steps 1–4 above, not a plain mean of the printed figures. The Average Calculator serves as an average of percentages calculator in the equal-bases case (paste the percentages as numbers), and its weighted mode takes the parts-and-bases method directly when they differ: enter each part as a value and each base as its weight’s partner — or simply pool the totals yourself and divide once, as the table prescribes.

Sources: the methods are the standard arithmetic of weighted means applied to ratios; every figure — including both pooled checks and the savings-rate example — was recomputed independently for this article. Part of the CalcFolk averages series, alongside the Average Calculator.