The ecological fallacy: group data does not describe a person

The ecological fallacy is the mistake of assuming that a pattern found in group-level data also describes the individuals who make up that group. A country, a state, or a school can show one trend in its averages while the people inside it show something quite different, or even the reverse.

The term traces to sociologist William Robinson's 1950 study of the 1930 United States census. Comparing states, Robinson found that states with a higher share of foreign-born residents had a higher average literacy rate, a positive correlation. But at the individual level, foreign-born residents were themselves somewhat more likely to be illiterate than native-born residents, the opposite direction. Immigrants had settled disproportionately in states that were already more literate overall, which produced the reversal.

Two traps follow. First, a correlation between group averages says nothing certain about the correlation between the same traits in individuals, and the two can even point in opposite directions, as Robinson showed. Second, the fallacy is easy to miss because group-level data, such as a country's income and its health outcomes, are often the only numbers available, tempting a writer to describe individuals with them anyway.

When a claim rests on comparing places, such as countries, states, or school districts, rather than people, ask whether the pattern has actually been checked at the individual level too. A state, city, or country's average tells you about that place, not about any one resident in it. Pair a group-level statistic with individual-level data before using it to explain a single person's behavior or risk.