Thanks for linking to the paper, I confess I hadn't gotten that far in the article. I would say they "aimed for" rather than achieved identification (condition

Thanks for linking to the paper, I confess I hadn't gotten that far in the article. I would say they "aimed for" rather than achieved identification (conditional ignorability) here...a bunch of backdoor baths come to mind (likability, for one: these people got clerkships in part because they got the best letters of recommendation, which they got by impressing the right professors, which they got (I imagine) by being likable). They kind of address this here: > Rather, whether a former clerk succeeds is predicated upon having expertise about the justice whose vote she seeks to capture. If quality was a confounding variable, then we would expect to find that former law clerks, by virtue of being the best and the brightest, always outperform non-clerks. They do not. This is interesting, but effects only observed in a subset of the data -> complexity penalty! How much this bothers you is gonna vary. I'm a stickler for identification, so I view this paper's results as more tentative than you might. (Regarding the journal outlet: it's an interesting empirical finding, but is it really a landmark in terms of methods or findings? I don't think so. And if they were aiming for AER, maybe they should have included a structural model ;))