The Ecological Fallacy
The term ‘Ecological fallacy’ is used in Epidemiology and some other disciplines (such as Sociology) to refer to improperly inferring a causal association (or lack of association) at an individual-level based on a group-level relationship. This use of the word ecological is at odds with the alternative use of the word in the discipline of Ecology. Ecological methods from Ecology are inherently multi-scaled and address precisely the issue of this cross-scale inferential fallacy.
I argue that a broader understanding of ecological methods by non-Ecologists would be a start towards better understanding between the disciplines. The inclusion of real ecological methods in the other disciplines will also assist research to better understand the causes and effects of climate and climate change which are important gaps in knowledge needed to enable mitigation and adaptation of our society to future climate change.
Sociology
To clear up this confusion it is necessary to go back to the source of the use of the term ‘ecological’ by the urban sociologist Amos Hawley who used the term ‘human ecology’ in the 1950s to build on the theoretical traditions of Robert Parke and E. W. Burgess at the University of Chicago in the 1920s on the structure and development of cities. His influence in the Social Sciences led directly to the development a tradition of social human ecology, one largely without a bio-physical environment. The usage of ‘ecological’ to mean multivariate studies of complex systems stems from the sociological tradition.
Ecology
The term ‘ecology’ used in the Ecology discipline dates back to the 1870s, coined by German zoologist Ernst Haeckel (1834-1919) as Oekologie from Greek ‘oikos’ for house, dwelling place, habitation and ‘logia’ study of, then at the turn of the century the term became spelt ‘oecology’, which grew into ‘ecology’ in the early part of the 1900s.
Epidemiology
In epidemiology the term ‘ecological study’ is used to refer to studies where observations are taken at the level of a group (such as a country, school, or hospital) rather than at the individual (such as patient) level. It is well known though that when risk factors and outcomes are measured at an aggregate level, the relationship between the group-level variables may be different than the relationship between variables measured at the individual level. An often cited example used to illustrate the issue involved a 19th century study which found higher suicide rates within Prussian provinces that had higher proportions of Protestant residents. The conclusion that Protestant individuals (rather than Catholic individuals) were more likely to commit suicide cannot be inferred based on the observed association among the provinces. One possible scenario is that Catholic residents within the largely Protestant provinces had the high suicide rates, resulting in a positive association between percent Protestant and suicide rate 8. Extrapolation of aggregate results to individuals is a mistake in logic 9 which can lead to a potentially misleading conclusion 10.
Because of this limitation ‘ecologic studies’ are often scorned in epidemiology as inferior and only useful for exploratory or hypothesis-generating studies rather than as confirmatory. I argue to the contrary that there is the potential for a revolution in our ability to understand causal influences operating at multiple scales of space and time if we were to conduct truly ‘ecological studies’.
Quantitative Geography
There is also a closely related concept that should be noted. That of the Modifiable Areal Unit Problem (MAUP) as discussed in quantitative geography. There is a large amount of geographical literature on the MAUP. In this problem domain areal units (also called zones, regions, areas or polygons) inherently pose three problems to a researcher attempting to infer causal associations: scale, zonal and temporal:
- Scale; this issues is evident in the example above where phenomena investigated using data viewed at one scale may appear quite different (even opposite) using data aggregated at a different scale.
- Zonal; the zonal problem appears where phenomena investigated using data viewed using two sets of differing areas at a single scale can differ.
- Temporal; a third problem arises when analyzing data on modifiable areas when people keep modifying them by redrawing the boundaries over time.
In Conclusion
The term ‘Ecological Fallacy’ is itself a fallacy and non-Ecologists should be made aware of the existence of alternative ecologic methods from Ecology. This would be a start towards better understanding between the disciplines and enhance our abilities to mitigate and adapt to climate change.
