What Lott’s coding errors mean

Several people have commented on the latest developments. Atrios has resolved that Lott is a liar and a fraud. Kevin Drum has his usual nice summary. Jesse Taylor isn’t really interested because he believes that Lott has already been discredited. Julian Sanchez and Chris Lawrence are reluctant to draw conclusions yet.

I think we have, however, enough information to draw some conclusions. There really were coding errors in the data Lott used for his NAS panel presentation. Mustard seems to have conceded this when he withdrew a graph based on that data. And there really were coding errors in the data for Lott’s Confirming “More Guns, Less Crime” paper. Lott responding to the claim that there were such coding errors by withdrawing his name seems to be a concession on that issue.

Were the coding errors deliberate, or were they just the product of inexcusable carelessness? Lott’s behaviour in the survey affair makes it seem quite plausible that he deliberately cooked the data, but it really doesn’t matter that much. What researcher will take Lott’s claims seriously now?

And while I think that Ayres and Donohue have thoroughly demolished the “more guns, less crime” hypothesis, you don’t have to take my word for it, or even try to understand any of their statistics. Lott had plenty of time to come up with a response and he ended up taking his name off it. Maybe he can come up with something at some later time, but it seems that he has conceded defeat for now.