Ecological Fallacy, explained.
The ecological fallacy is an unsupported inference about individual relationships from relationships observed in aggregate groups.
Why it happens
Two districts differ in library access and passing rates. A detailed count ledger reveals equal pass rates for access and non-access groups within each district despite the positive aggregate association.
The ecological fallacy is an unsupported inference about individual relationships from relationships observed in aggregate groups.
Read the result
The district summaries describe groups. The detailed cells are required before making a claim about individuals within either district.
A worked example
Unpack a district-level association
Compare district A’s eighty-percent access and seventy-percent pass rate with B’s twenty and thirty percent.
Reveal the access-by-passing cells separately within each district.
Observe equal within-district pass rates for access and non-access groups without interpreting this example as a causal finding about libraries.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
Two districts differ in library access and passing rates. A detailed count ledger reveals equal pass rates for access and non-access groups within each district despite the positive aggregate association.
Where this idea is useful
A practical use
Check the level of analysis before attributing a neighbourhood trend to every resident or a programme’s causal effect.
A common misconception
“A positive group-level association must hold within each group.”
Differences between groups can generate an aggregate pattern without the corresponding within-group relationship; the level of analysis matters.
What this explanation leaves out
- Counts are stipulated illustrations, not evidence about libraries. Aggregate data can be informative with appropriate assumptions, but those assumptions are not supplied by a correlation alone.
Does this prove aggregate research is useless?
No. Aggregate data can answer group-level questions and support other inferences with justified assumptions; the unsupported level change is the problem.
At which level were the observations collected and at which level is the conclusion being stated?
Associated thinkers
Associations marked provisional are awaiting source review.
Further reading
Explore the original research or the teaching reference behind this experiment.