Step 1: Speak at Duke.
Step 2: Make sure that Brendan Nyhan is in the audience.
Step 3: Make sure that Kevin Drum reads Brendan Nyhan and cites your article.
I only wish that Kevin Drum would’ve linked to this pretty picture which I tihnk tells the story so cleanly:
Anyway, it’s nice to get these ideas out there. To end on a technical note: Drum gives us a bit too much credit when he writes,
The chart above [Figure 13 of our paper] shows how clear this distinction is, and the authors report that it’s become clearer over time. It’s rare to get such a clean regression line in social science data like this, which makes their results pretty provocative.
It’s not quite so clean as all that, however: the points on that graph are estimates from the multilevel model, not raw data, and the model itself pulls the points toward a straight line (to the extent that this line is supported by the data). See Footnote 9 on page 19 for an elaboration of this point. Just to be clear: I do believe this pattern is striking–and, amazingly enough, we discovered it in our analysis, we did not go looking for it. But the super-crispness of the line is somewhat a product of the modeling.