The role of empirical work
Einav and Levin write:
Hamermesh recently reviewed publications from 1963 to 2011 in top economics journals. Until the mid-1980s, the majority of papers were theoretical; the remainder relied mainly on “ready- made” data from government statistics or surveys. Since then, the share of empirical papers in top journals has climbed to more than 70%.
Isn’t that remarkable? I certainly was under the wrong impression when I was a Ph.D. student in Berkeley and Mannheim and thought that it’s all about theory and methods. Where does this come from? Maybe it was because one tends to see so much theory in the first year of a full-blown Ph.D. program, which is full of core courses in Micro, Macro and Econometrics, covering what is the foundation to doing good economic research. In any case, my advice to Ph.D. students would be to strongly consider working with real data, as soon as possible. There is certainly room for theoretical and methodological contributions, but this should not mean that one never touches data. At least in theory 😉 everybody should be able to do an empirical analysis. And for this, one has to practice early on. Even if one wants to do econometric theory in the end. But even then one should know what one is talking about. Or would you trust somebody who talks about cooking but never cooks himself? OK, I admit, this goes a bit too far.
After having said this let me speculate a bit. My personal feeling is that one of the next big things and maybe a good topic for a PhD could be to combine structual econometrics with some of the methods that are now used and developed in data science (see the Einav and Levin article along with Varian‘s nice piece). In Tilburg, for instance, we have a field course in big data, by the way, and another sequence in structural econometrics (empirical IO).