Ecological Fallacy

The ecological fallacy is a type of faulty reasoning that sometimes is made in the interpretation of results that come from the analysis of aggregate data. This mistake occurs when data that exist at a group or aggregate level are analyzed and interpretations are then made (generalized) as though they automatically apply at the level of the individuals who make up those groups. For example, if a researcher used zip code level census data to determine that the proportion of women in the labor force was inversely correlated with the prevalence of mobile homes in that zip code, it does not necessarily follow that women who live in mobile homes are less likely to be employed than are women who do not live in mobile homes.

It is possible that the same relationship between employment and type of housing exists at the level of individuals, but just because it was found to exist at the aggregate level does not assure that it holds at the individual level.

The ecological fallacy can come into play for survey researchers who merge aggregate level data onto their survey data sets—original data that are gathered in a survey from the individual respondents. For example, if a survey of adult residents in a metropolitan area were conducted about race relations, the researchers may want to enhance their analyses by merging zip code or block group census data onto each case in the data set; for example, merging area-level variables such as the percentage of residents in the zip code or block group who are white and the percentage who are black with each survey respondent's individual-level data. These variables can serve many purposes, including being used as statistical controls. They also can allow the researchers to generate new variables by using both the individual-level data gathered in the survey and the aggregate area-level data merged onto the data set; for example, creating a new variable that indicates whether the respondent lives in a zip code or block group in which her or his own race is the majority race. There is nothing inherently wrong with doing any of this, and it does not constitute an instance of committing the ecological fallacy.

Instead, the problem of committing the ecological fallacy occurs when researchers go beyond the precision and applicability of their data to draw conclusions that the data simply do not justify. If the findings are based only on variables that exist at the aggregate level, then no conclusions should be generalized to the individual level. That is not to say the researchers cannot speculate that the same relationships may exist at the individual level. But that is as far as the researchers should go, and it should be labeled explicitly as speculation due to the possibility of the ecological fallacy. If a researcher believes it is important to determine whether the relationship holds at the individual level, it is her or his responsibility to investigate it by conducting a new study that gathers appropriate data at the individual level or by conducting secondary analyses of existing individual-level data.

...