A fallacy is an error in reasoning. The ecological fallacy occurs when conclusions are made about individuals based on group data.
Any group whether it be St. Louisans, community college graduates, immigrants, BMW drivers, people who own fish, etc. will have a range of individual traits. Just because the group has a characteristic doesn’t mean that each individual also has that trait. A few examples will make the fallacy clear.
Example One: The classic example of the ecological fallacy was published in 1950 by W.S. Robinson. Data showed that in 1930 those states with the highest proportion of foreign-born residents also had the highest literacy rate. That data made it seem like foreign-born people were more literate in the English language than the native-born. Upon first hearing this your brain can start jumping to all sorts of explanations: maybe the foreign-born work harder to learn English to assimilate; maybe the foreign-born have a greater work ethic.
Actually, if you dig into the individual data, the foreign-born actually have lower levels of literacy than the native-born. The fallacy occurs because “foreign-born” are individuals within the larger group of “residents of a state.” There are many individuals that make up the residents of each state with the foreign-born being a small proportion. The reasons for state literacy are based on many factors and many individuals. Thus, ascribing group characteristics, like literacy of a state, to individuals in the group, such as being foreign-born, is an error.
Example Two: Here’s a more simple example. Studies show that millennials don’t like to work as many hours per week as older generations. You interview Steve, a millennial, for a job at your company. Should you assume that Steve will be less willing to work long hours than older generation employees?
No. There are tens of millions of millennials and the conclusion that they are less willing to work long hours is a conclusion based on the aggregate data from the large group. The individual members of the group have large variability and the conclusion for the group should not be applied to Steve.
Example Three: Studies show that high school students of Asian descent perform better on mathematics exams than non-Asian students. If you meet two students, Eric, who is white, and Emmie, whose parents emigrated to the U.S. from Japan, should you assume Emmie is better at math than Eric?
No. Again the conclusions based on mathematics scores are based on millions of students with each group having a very large range of results. For instance, if Eric is above the white average and Emmie is below the Asian average it is likely that Eric scores better than Emmie on math tests.
In conclusion, be wary of ascribing group characteristics to individuals because the group characteristics are usually derived from a wide range of underlying individual attributes.
Two related interesting points are: (1) the differences within groups are usually greater than the differences between groups and (2) correlations are not transitive. These two points speak more in-depth about why attributing group characteristics to individuals from the group is not appropriate.