Will Rogers Phenomenon — when the Okies moved out, both states got smarter

Dr. R. Düsing · Osnabrück University
Will Rogers Phenomenon — Help
The joke

US comedian Will Rogers quipped: "When the Okies left Oklahoma and moved to California, they raised the average IQ of both states." 😄 Tongue-in-cheek — but statistically entirely correct, and that's exactly what this tool shows. (No real claim about anyone, purely a statistics gag.)

Does / does NOT

Does: shows how simply reclassifying cases between two groups can raise the mean of both groups at once — without a single individual value changing.
Does NOT: it is not a real effect, not a learning effect, not measurement error. The overall average stays exactly the same; only a label moves.

The mechanism

A case with value w moves from group A to B. Both means rise exactly when

x̄_B < w < x̄_A

The migrant is below average in A (leaving raises x̄_A) and simultaneously above average in B (arriving raises x̄_B). Here: Oklahoma has the higher average, California the lower one; the migrants sit in between — below Oklahoma's average but above California's.

The overall average doesn't change because no value itself changes — only the assignment. The sum and count of all values stay the same.

The controls

Willing-to-move Okies: how many Oklahoma residents are willing to relocate at all (outlined in purple in the plot). ▸ Start wave: plays the move automatically — the Okies migrate to California one by one (lowest IQ first), both mean lines visibly shift right. …or by hand: the same move, controlled manually (0% = no one has moved yet, 100% = everyone).

Where it gets serious

The same effect is called stage migration in medicine: better diagnostics reclassify cases into different disease stages, making the statistics of every stage look better — without any real progress (Feinstein 1985). Rankings, grading tiers, or league tables are just as susceptible whenever cases get regrouped.

What to do?

Don't compare group-specific means across different classification rules. Instead, look at the overall group (immune to reclassification) or apply the same classification consistently. Get suspicious when "every subgroup improves" but the overall value stays flat.

References

Feinstein, A. R., Sosin, D. M. & Wells, C. K. (1985). The Will Rogers phenomenon. New England Journal of Medicine, 312(25), 1604–1608.

📋 Example — Okies Move to California
Will Rogers' quip: "When the Okies left Oklahoma and moved to California, they raised the average IQ of both states." 😄 The migrants are below Oklahoma's average but above California's average — they sit exactly in between. Question first: what happens to both states' averages when they move — do they rise, fall, or a mix? Slide the migration wave up (or click "▸ Start wave") and watch.
① Average IQ by State
Oklahoma 🤠
—
—
California 🏄
—
—
USA overall
—
unchanged by the move
Moved
0
of 12 willing to move
No one has moved yet — slide the migration wave up.
② Each Point = One Person · Position = IQ
top lane = Oklahoma · bottom lane = California · thick line = mean now · pale line = mean at start
Concepts
What is the Will Rogers phenomenon?
If you move a case from group A to B whose value lies below A's average but above B's average, the mean of both groups rises — even though no individual value changes. The overall average stays constant.
Why does the overall average stay the same?
A move only changes a label (place of residence), not a measured value (IQ). The sum and count of all values stay identical → the overall mean is immune to reclassification. Only the split shifts.
The serious side: stage migration
In medicine, better diagnostics reclassify cases into different disease stages — every stage looks better afterward, without any real progress (Feinstein 1985). Rankings and grading tiers are just as susceptible.
How to spot it / what to do?
Suspicious: "every subgroup improves," but the overall value stagnates. Solution: compare the overall group, or apply the same classification consistently. → Simpson's Paradox → Regression to the Mean