When Averages Lie: Understanding Simpson’s Paradox

Imagine you’re comparing two treatments for a disease. Treatment A works better for men and women separately—but when you look at the combined data, it suddenly appears worse. Cue a head scratch: how can pooling results flip the story? Welcome to Simpson’s Paradox, a data analysis quirk where averages can lie unless you pay close attention to hidden factors.

Where did this come from?

The paradox is named after British statistician Edward H. Simpson, who described it in a 1951 paper. But the phenomenon was spotted earlier in 1899 by Karl Pearson when analyzing hospital recovery rates. A famous cultural moment came in 1973 when UC Berkeley’s graduate admissions data seemed to show bias against women—until analysts discovered that departments with low admission rates had more female applicants, revealing a classic Simpson’s Paradox twist.

Where you’ll see this in real life

• Medical studies: A treatment may look beneficial within age groups but harmful overall if one group skews the numbers. • Sports stats: A baseball player’s batting average can appear higher when split by home and away games yet lower when totals combine. • Business decisions: Aggregated sales figures might suggest one region underperforms, hiding strong local markets. • Public policy: Education or crime rates can flip signs when factoring in socioeconomic or geographic variables.

A common pitfall

The trap is ignoring “lurking” variables—hidden factors that change the story when you pool data. When teachers or students rush to conclusions based on overall averages, they might miss the real insights lurking in subgroups. Always ask: What’s been combined, and what might be hiding beneath the surface?