Labor economists love using instrumental variables – like your distance to the nearest college – to estimate the education premium.  Heckman, Lochner, and Todd point out a recurring problem with this approach:

[M]ost of the candidates for instrumental variables in the literature are also correlated with cognitive ability. Therefore, in data sets where cognitive ability is not available most of these variables are not valid instruments since they violate the crucial IV assumption of independence. Since few data sets have measures of cognitive ability, this finding calls into question much of the IV literature. Notice that the local unemployment rate is not strongly correlated with AFQT. However, it is only weakly correlated with college attendance.

Check it out:

heckman.jpg