Barbara Bergmann (emeritus professor at American University) suggests we need to reorient economic empiricism. Here's the beginning of her column on the subject:
"The New York Times has reported a discovery by biologists studying how bottlenose dolphins hunt fish in groups. They found that individual dolphins consistently took specialized roles: one particular dolphin always served as herder, and the others served as barriers to the fishes’ escape. The scientists discovered this by watching two groups of dolphins during 120 feeding sessions.Probably a slight overstatement in that last sentence, but the comparison of biology and economics is still quite stark. To a large extent it's a stark difference for good reasons. We have much better data available than biologists that don't specialize in the particular species of highly evolved primate that interests us. Why go for a small sample field study when there's plenty of work to be done on the ACS?
Those two dolphin groups probably received more close attention in this single study than any of the human groups known as business establishments had received from professional economists in the last 200 years."
But clearly there are questions better answered going out in the field, and economics departments don't really prepare you at all to do that work.
One of the things I like a lot about the Urban Institute is that they combine field work with data work. A lot of the field work is headed up by public policy or sociology PhDs, but the economists actually participate in it too. It's a model that more economists should follow.
Field work has costs too, of course. As we know from work with datasets, it's very hard to parse out the truth in observational data. The methods we've developed to do that often require careful model design and lots of sample size. That's just not there in field data. But the point is not to just do one or the other - the point is to rely on the strengths of each approach. Moreover, economists that are trained to worry about model identification are exactly who you want to have working on field studies because they know the pit-falls (and vice versa - you want people trained to do field studies looking over the modeling).