Clearly the most useful subfield of economics for the world right now is development economics.
Also quite clearly, econometrics and basic undergraduate micro theory is more than adequate to conduct Duflo-type research. Should graduate programs shift their focus to recruit students with more statistics (not pure math), and make curriculums more empirical, as opposed to mathematical.
My physics colleague once remarked that economics is more "mathematical" than physics. Not that it uses more mathematics, but economists seem to think that more formality is helpful/meaningful.
We know economic theory has about the same chance of finding a "grand unified theory" of economic behavior as string theory has of the universe. So why are not interested in doing economics that is actually useful to policymakers, not just a series of increasingly futile attempts to find consistency in an entropic system.