The Ecological Fallacy and Aggregation Bias

Suppose sectors with a higher average dividend yield also show higher average returns across a decade of data. It's tempting to conclude that high-dividend stocks individually outperform, but that's exactly the leap the ecological fallacy warns against. A relationship measured on aggregated groups (sectors, countries, funds) doesn't automatically apply to the individuals that make up those groups; the sectors driving the pattern might contain a mix of a few spectacular high-yield winners and many mediocre high-yield losers, while the individual-level relationship between yield and return is flat or even negative.

This happens because aggregation averages away exactly the variation you'd need to see to check the individual-level claim. Two groups can have identical averages while containing wildly different underlying distributions, a group of moderate performers and a group split between big winners and big losers can post the same mean, but only one of them contains any individual worth picking out. Backtests that only ever report portfolio-level or sector-level statistics are especially exposed to this: a factor that "works" in aggregated decile returns might be driven entirely by a handful of names, with no reliable effect on the typical stock in the decile.

The guard against it is simple in principle, if not always in data availability: check the claim at the level you actually intend to act on, not one level of aggregation above it.