In some cases it is not possible to determine the effect of a policy since whether or not an individual is treated depends in part on decisions taken by them in relation to unobservable variables Take, for example, their reasons, ability, intelligence and other individual attributes.

In these cases an external factor that exerts a strong influence on the selection process but at the same time has no effect on the result variable must be found.

By using this external factor, commonly known as the instrumental variable, it is possible to break down the effect of the treatment into two components: one determined by the external factor itself and the other by the unobservable variables. At this point only the part determined by the external factor will be used as a treatment indicator, thus eliminating the influence of the unobservable variables.

The main limitation of this method lies in the difficulty of identifying or constructing an appropriate instrumental variable.