ABSTRACT
Subjects may be allowed to vary on extraneous variables if it is believed that their variation does not have any relationship to the causal variables of the study. Any influence on the dependent variables of such variation by extraneous variables will then enter as components of the disturbances, but the causal relations will be unaffected. In such cases the causal relations will be estimable without bias. But even in the cases where some of the variation in individuals is correlated with causes of these variables, it may be possible to still estimate the causal relations as long as they are endogenous and there are exogenous variables, this time known as instrumental variables, that are causes of the endogenous causes, but not directly of their effects. These instrumental variables must be independent of the disturbances on these endogenous effect variables.
