Ecological and individualistic fallacies in health disparities research

Abstract

Ecological fallacy, defined by Robinson in 1950 as incorrect inferences about individuals based on characteristics and associations observed among groups ( 1), is a well-recognized concept in epidemiology and statistics. Research has shown when aggregate values for variables of interest (eg, median area-level income) were used as proxies for individual-level variables (eg, household income), covariates estimated from regression models may be biased ( 2, 3) and the sign of regression coefficients could change ( 4). Policy makers and health-care providers who rely on these estimations could inadvertently draw the wrong conclusions or target the wrong group for interventions.