|
|
| The essence of the identication problem concerns our ability to test theories empirically. Most theories concern predictions about the relationship between variables when other factors are held constant (this is called the ceteris paribus assumption). For example, we have a theory about how a variable X affects Y. We get data on X and Y to test the theory. The problem is that, outside of our model, Z also affects Y. |
| If we could conduct experiments, we would design one in which Z is constant. But there are two challenges here. First, we would have to know that Z also affects Y in order to be aware that we should design an experiment to deal with it. Second, it is usually not possible to design experiments in economics. |
| The transparencies and readings give several examples of the identification problem. The simplest one is the supply and demand example, which I won't repeat here. Another version of the identification problem has its own name: the sample selection problem. |