Is the “Ecological Fallacy” a Fallacy?

Abstract

Ecological studies of health effects due to agent exposure are generally considered to be a blunt instrument of scientific investigation, unfit to determine the “true” exposure-effect relationship for an agent. Based on this widely accepted tenet, ecological studies of the correlation between the local air concentration of radon and the local lung cancer mortality as measured by Cohen have been criticized as being subject to the “Ecological Fallacy” and thus producing invalid risk data. Here we discuss the data that a risk assessment needs as a minimum requirement for making a valid risk estimate. The examination of these data and a “thought experiment” show that it is Cohen's raw ecological data, uncorrected for population characteristic factors, which are the proper data for a risk assessment. Consequently, the “true” exposure-effect relationship is less and less important the more population characteristic factors are identified and the larger they are. This reversal of the usual argument is due to our approach: Here, the prediction of the health effects in an exposed population is of primary importance, not the shape of the “true” exposure-effect relationship. The results derived in this paper hold for ecological studies of any agent causing any health or other effect.