Frances Wooley has a great post on modern econometrics. Here's the key point:
"Any exogeneous phenomenon that is closely related to an economic outcome of interest is usually thoroughly understood. Researchers will have had numerous opportunities to observe cause-and-effect relationships. Hail destroys crops and causes the price of wheat to rise. Cold winters increase the demand for fuel and cause the price of heating oil to rise. The low-hanging fruit have been picked, processed, and made into jam.
This mean that, in practice, researchers face an exogeneity-plausibility trade-off -- hoping to find truly exogeneous explanatory factors, they have turned to less and less plausible explanations...
The problem is that exogeneity/endogeneity is easier to assess than plausibility. So a paper with an exogeneous but wacky explanatory variable has a reasonable chance of being published. A paper with a potentially endogeneous but sensible explanatory variable faces more challenges."
The problem is worse than this, I think. Economists love things that are counterintuitive and quirky, which means they get excited about strange instruments. My view is that nobody ever trusts anyone else's instruments anyway. Most people are pretty skeptical. So if they don't buy it anyway, they at least want to be entertained by a clever paper.
Wooley also links to the Bound, Jaeger, and Baker paper that I mention on here a lot when I talk about IV models. It's one of the most widely known take-downs of Angrist and Krueger's returns to education paper. I think probably the single best influence on me as an economist so far was to have Jaeger as both an econometrics and labor professor at William and Mary - it's rare to get someone honest enough to be (1.) worried about endogeneity, but also (2.) skeptical of IV approaches. Usually people that worry a lot about endogeneity go whole-hog on IV because it seems to offer a solution. And it certainly can offer a solution, but a lot of IV approaches are pretty dicey and you have to be cognizant of that. I prefer instruments that come in the form of sharp discontinuities that are hard to anticipate or control. They are much cleaner, in my opinion. One of the things that has concerned me about the way macroeconomists think is that they generally don't seem to be as worried about these identification problems when they do empirical work. Unfortunately, even if you work hard at that (like Barro or Romer), there's very little material to work with.
UPDATE: Does everyone know what I mean when I talk about instrumental variable approaches? I just realized this might not be entirely clear. If not, I can do an explanatory post tomorrow.