Social factors and health: the causation-selectionissue revisited.

Since social scientists rarely have access toexperimental data, they rely heavily upon observational studies. As aconsequence, their attempts to make causal inferences about the effects ofsocial factors--such as occupation or marital status--on health are plagued bypotential selection problems. Some researchers have addressed thisselection-causation problem on the basis of the presence or absence of aparticular aggregate pattern of health status. The rationale underlying thisapproach derives from the investigators' hypotheses that the presence ofselection would lead to a particular type of pattern that is distinct from thepattern that would result in the absence of selection. Although intuitivelyappealing, this strategy appears to be seriously flawed. The essential weaknessis that the range of patterns that can result from selection is often muchbroader than researchers have speculated.