Raj Chetty: evaluation

Transcript context

…But those both sound replicable, right? What’s the non-replicable asset? What hopefully our contribution and scale is, is showing how you can take those large datasets and not get lost in them, and bring out the key lessons that are relevant for thinking about these classic questions. It’s very easy — students often have this reaction, that all I need to do is get access to this big dataset, and then I’m going to be all set for my thesis. And what you end up finding is that that is often not the case. It’s very easy to write a paper that is not that good, even with cutting-edge data and modern techniques. So one of the things that I try to do — and the easiest way to see this is if you internally, within our research group, see the iterations of the papers we’ve been working on — where we start out is often very far away from the papers that people see as the finished product. We work hard to try to write a paper that ex post seems extremely simple: “Oh, it’s obvious that that’s the set of calculations you should have done.” Let me give you an example, a concrete example, in the context of the paper you just mentioned — the one that came out after that Twitter comment you mentioned, where we look at the fading American dream and look at how children’s chances of doing better than their parents have changed over time. Well, it turns out, when you dig into the details of that analysis, in order to know if a child is doing better than their parents, you would think you need to have a dataset where you see both the parents’ income and the kid’s income. It turns out you don’t have that data in the US going back to 1940, which is when we wanted to start our analysis. In fact, you only have that data starting really for kids born in the 1980s. So the key conceptual idea of that paper was to figure out a way, even though you could not link parents and kids, a statistical method of determining the fraction of kids doing better than their parents despite that data limitation. Often our conceptual advances are in the context of taking a big question, using the best data that we can possibly get, but then closing that gap and trying to find creative methods and creative answers to those old questions, and executing in the best way possible. One thing I take pride in is the conclusions we reach and the statements we make, we’re quite confident in them. We take pride in the fact that subsequent studies do not end up showing, “Oh, there was some assumption made that ends up being questionable,” or “The data is not quite up to par,” etc. We’re very careful on all those steps. If I’m trying to model the Raj Chetty production function and I described it as such, it’s a multiplicative model, so there is getting the data, but that’s not the key point.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.