Describes a fallacious inference in which a statistical characteristic is inadmissibly transferred to a level of aggregation lower than that which it actually describes.
This may, for example, take the form of a (faulty) statistical syllogism, as in this example:
Most people from Scandinavia are blond.
Björn comes from Scandinavia.
Therefore: Björn is blond.
Of course, we don’t know whether Björn isn’t one of the few none-blond Scandinavians, so we can not draw such a conclusion.
Statistical indicators are always simplifications of the underlying data. When complex issues are reduced to one or a few key figures, they are easier to understand, communicate and compare. At the same time, however, we also lose information, especially about the data available on individuals.
For example, a city’s crime rate can be a useful metric for tracking developments over time or to compare it with other cities. It is however not suitable for infering information about individuals or subgroups (e.g. neighbourhoods) from this city.
The fallacy consists in projecting such a metric, which refers to a higher level of aggregation, onto a lower level. In this instance: by assuming that information we have about the city as a whole (the higher aggregation level) allows us to infer information about a neighbourhood (a much lower aggregation level), or even about individuals (the lowest possible aggregation level) from that city.
Specific to this fallacy is that it relates to statistical metrics. In contrast, the similar mereological fallacy applies to the relationship between a system as a whole and its (functional) components. Furthermore, the fallacy of division refers specifically to transfers resulting from the phenomenon of emergence.
It is not always easy to distinguish between all of these. There are often instances that could fall into several categories. In some sources, these fallacies are thus all simply treated as synonyms for each other.
This article is still in preparation and thus incomplete.