The mean moves, the median does not: one billionaire changes the average income and not the typical one

01The mistake

Add a very large value to a data set and ask what happens to the mean and the median. Students say both increase, by similar amounts, because both are “the middle” and a bigger data set with a bigger number should have a bigger middle. In fact the mean can move enormously while the median moves by one position, or not at all.

The tell is asking which measure to report rather than how to compute one. Computation questions are answered correctly by procedure; the reporting question requires knowing what each measure is sensitive to. A student who says it does not matter, or picks one at random, has the two fused whatever their arithmetic looks like.

It drives the skew questions too. In a right-skewed distribution the mean sits above the median, and students who believe the two track together cannot predict which is larger, so every question about skew becomes a coin flip. Worth noticing that this is not a separate topic — it is this misconception being assessed in graphical form.

The related error is treating the median as an average of the middle values in all cases, or failing to order the data first. That is U4-SM5, and it travels with this one: a student who has not grasped that the median is defined by position rather than by value tends to make both errors.

02Why it makes sense to the student

They are introduced together, as a list. “Measures of center: mean, median, mode.” A list implies interchangeable members, and the lesson usually spends its time on how to compute each rather than on what distinguishes them. Nothing in the presentation says these will ever disagree.

On the data sets students practise with, they usually agree. Textbook sets are small and roughly symmetric, so the mean and median come out close, and dozens of exercises confirm that either one works. The cases where they diverge are the cases that get one problem at the end of the section.

“Average” in everyday speech means both, and neither precisely. People say average height, average score, average income, and rarely specify. Income is the case where the choice matters most and where the ambiguity is most often exploited, which students have usually never had pointed out.

And the median's definition sounds like a computation when it is really a rule about position. Students learn “put them in order and take the middle one,” and the ordering step feels like setup rather than like the source of the property that makes the median resistant.

03The correction

Anchor on what each one uses. The mean is computed from every value, so changing any value changes it. The median is determined by position, so it depends on the order of the values and not on how extreme they are. That single sentence generates every consequence in the topic.

Demonstrate with the smallest possible example. Take $\{1, 2, 3, 4, 5\}$: mean 3, median 3. Change the 5 to 100: mean becomes 22, median stays 3. One value moved, the mean moved by 19, the median did not move at all. Students find it more convincing to watch the median refuse to move than to hear that it is resistant.

Then push it to the real case, because it is the one they will remember. Ask which measure a town should report for typical income if one resident is a billionaire. The mean is technically correct and useless; the median is what anyone actually wants to know. This is also where the word “average” being ambiguous becomes a point about truthful reporting rather than a technicality.

Connect it to skew directly so the two topics fuse the right way round. In a right-skewed distribution, the long tail of large values pulls the mean toward it while the median stays near the bulk of the data, so mean $>$ median. Left-skewed reverses it. Students who hold the sensitivity principle can derive this instead of memorising which way the inequality points.

A useful classroom test: “A data set of nine values has its largest value increased from 20 to 2,000. What happens to the mean? To the median?” The answers are “increases a great deal” and “does not change.” A student who moves both, or neither, has them fused, and the direction of the error tells you which one they are anchored on.

04A sample question

05What each wrong answer reveals

• A The two measures fused. The most common wrong answer. The student reasons that a much larger value makes the data set larger overall, so any measure of center should rise. They are right about the mean and have applied the same logic to a measure that does not work that way. The five-value demonstration is the correction — watching the median hold still does the work that an explanation does not.
• B Correct. The mean rises from 6 to 24. The median is the third value in order, which is 6 both before and after, since changing the largest value does not change which value sits in the middle.
• C The properties swapped. The student knows the two behave differently — genuinely more than A knows — and has assigned the sensitivity to the wrong one. Usually it means they have anchored on the ordering step and concluded that the median is the one that responds to the data. A quick fix, because the conceptual distinction already exists.
• D One value read as negligible. The student has treated a single change as too small to matter in a set of five, which is a reasonable instinct badly calibrated — one value out of five is 20% of the data. Uncommon, and it points at a shaky sense of how the mean is built rather than at the mean-median distinction. Have them compute both sums.

C is the answer worth being pleased about. That student has the concept — the two measures respond differently — and needs only to learn which is which, whereas A needs the distinction built from nothing. D is a separate problem about what a mean is. Grading these as one pile of wrong answers loses the fact that a third of them may be one sentence from correct.

06Try it in Mistake Master