People are not passive acceptors of threats to health: endogeneity and its consequences

The effect of behaviour on health is a major area of contemporary epidemiological enquiry. Most epidemiological studies of the effect of behaviour on health assume that the levels of the behaviour-related variables are determined by factors other than those under study. However, in many instances, obvious examples are breastfeeding and smoking, not only do behaviours affect health but, conversely, individuals take into account their (observable and non-observable) health conditions when making behavioural decisions. In models which allow for the joint determination of health and behaviour, both health and behavioural variables are 'endogenous', that is, determined by forces acting within the model. Through some simple didactic examples it is shown that estimates of the effect of behaviour on health are biased if endogeneity is ignored. Review of the small empirical literature on this subject shows perverse results, such as a negative relationship between the use of prenatal care and infant mortality, when endogeneity is ignored. Standard procedures for taking account of the effects of endogeneity are described briefly.