It would have been an excellent example in two of my classes. I'm referring to the graduate student who tried to replicate the Rogoff and Reinhart paper and found they had made a substantial mistake. The student was working on an assignment for an econometrics course. Here's part of the assignment:
The term paper assignment can be in one of two forms: a replication of one empirical econometric paper that interests you, or an original piece of econometric research. Given other demands over the course of the semester, we strongly recommend that you opt for the replication option, which should generally be less demanding than coming up with, estimating and writing an original research paper. At the same time, a replication can be a very illuminating experience . . .
Three comments:
1. Over the last dozen years I've taught undergraduate and graduate tools courses and have also asked the students to do a replication. In a single semester in which they have other courses and other things to do, I absolutely agree that originality should not be stressed. This is because, as should be obvious--but wasn't to my econometrics instructors--it's quite difficult to be original on a timetable. A replication is possibly "doable" and that's vital.
2. Another advantage of a replication is that any researcher must answer a fundamental question: will anyone else be interested? Replicating a published paper should usually answer that question in the affirmative.
3. It could be "a very illuminating experience". Indeed! I second that. My students have written authors of published papers for their data or simply to ask a couple of questions and have often been stonewalled. That's a harsh but useful introduction to the world of academic research.
As I often mention (i.e., whine), I've been stonewalled on data requests myself (was the CVM data collected by Exxon ever released?). Sharing data is always scary, even if global austerity policy or natural resource damage awards isn't/aren't on the line, because of the R/R problem. You always wonder what you may have done wrong in your spreadsheet (or, in my case, SAS program ... luckily for me fewer and fewer graduate students know SAS).
Me? ... I am always happy to share even when the authors are able to hit better journals with my data than I am myself (I should list two JEEMs on my vita under the heading "I wrote the survey but couldn't write the paper if my life depended on it").
Send me your data requests! But please don't go on Colbert if you spot a mistake.
