Do we need more field experiements and RCTs in doing empirical economics or do we need to rely on alternative methods and develop new econometric techniques? Duflo's Clark Medal has triggered some methodological debates among the development economists recently (see here and here). One possible alternative (as they cite Acemoglu's piece) is to rely more on structural models in which data for counterfactuals could easily be simulated rather than collected via field trials.
To me, this is just another around of methodology debate between the reduced form school and the structural school. Any deep issues aside, for graduate students, it might be a safer strategy if we can do both.