Well shoot, I still think Brad DeLong was very unfair to John Cochrane when he said that in his Cochrane’s WSJ op ed, Cochrane “draws this curve:” and then posts a curve that DELONG drew, not Cochrane. That’s the upfront, gripping part of the post, and when I first read it, I too was stunned that Cochrane had (allegedly) done that.
Later on in the post, DeLong posts computer code and says “Play with the R-code if you want to see how much a more flexible functional form wants to say that the U.S. has the optimal “Business Climate.” I confess I didn’t look very carefully at this. I thought DeLong was saying (I’m paraphrasing), “Even if you use other functional forms, you can’t make the ‘best-fit’ line go through the point Cochrane needs, if you’re using levels of per capita income on the y-axis.”
But alas, I should’ve looked more carefully. DeLong at that point (again, the bottom of his post) *does* finally switch to log income, which is what Cochrane actually used in his WSJ post. (Cochrane explains why he did that, in his blog response to DeLong.)
So, I still would’ve had a nice little blog post had I looked more carefully at what DeLong did with that code. To wit: Rather than simply reproduce the actual graph Cochrane used, instead DeLong (a) first tells us he drew a graph that he did not, and (b) then gropes around with computer code to make it work better (?), leading the reader to believe that DeLong is searching for a way to figure out where Cochrane could possibly be coming from.
So anyway, I took down the first post, because I had so badly misunderstood that component about the computer code that it was easier just to remove it. My apologies for the confusion I added.
Last thing: I realize it sounds like I’m making excuses, but guys, cut me some slack. I don’t read DeLong’s blog anymore, or Cochrane’s for that matter. The whole reason I even stumbled on this dispute was that I (foolishly) follow Noah Smith on twitter, and this is the tweet that alerted me to DeLong’s blog post:
So I went into this thinking–and Salim Furth confirmed–that DeLong had eviscerated Cochrane, both logically and with computer code. This is why I didn’t carefully parse DeLong’s computer code to see that far from writing code that “annihilates Cochrane’s main graph,” I think actually what happened is that DeLong’s code came closer to explaining Cochrane’s main graph (though still not sure what he was doing with it–I think Daniel Kuehn and Noah Smith came to opposite conclusions about what DeLong was demonstrating with that code).
What did we learn, kids?
#1) Never be really smug, because in case you make an innocent mistake you end up looking like an idiot. (I’m referring to myself in the previous post.)
#2) Economists are dangerous.
#3) I need to unfollow Noah Smith on Twitter. I’ve known this for a while but this is pushing me to act. I’ve hit rock bottom.
#4) I made Daniel Kuehn’s weekend.