Sunday, November 18, 2007

Manski and Identification

I am teaching Charles Manski's great book on Identification Problems in the Social Sciences right now (I must say that I find his writing much clearer than the class I had with him as a grad student).

One of this important points is that one cannot evaluate the counter factual or a persons' life. Suppose we find that children who come from in fact families graduate from high school at an 87 percent rate. Then without assuming a lot of structure, all we can know is if children in non-intact families could magically be transported to intact families, the children's graduate rate would lie between 0 and 100 percent. It is only if we assume that kids got randomly picked into intact and non-intact families that we can assert that all kids in in-tact families will graduate at the 87 percent clip. But this assumptions seems hardly reasonable. More on this tomorrow...

No comments: