The following (quoted with permission) is from an email to me from John Donohue. Donohue and his co-author had suggested — it turns out incorrectly — that Lott’s withdrawal of his name from the Stanford Law Review paper might represent a confession of error. Since that assertion has traveled around the blogosphere, Donohue thought it appropriate to acknowledge that it was not the case, and to reiterate his substantive challenge to Lott’s results:
Ian Ayres and I showed that virtually every regression in Lott’s reply to our paper in the Stanford Law Review was incorrect due to Lott’s coding errors. As a result, when he took his name off the paper, we assumed that this meant he was conceding that his coding errors undermined the results in his reply. We now understand that he “stands by” the erroneous regression results, and that he claims to have removed his name because of a dispute that he had with the editors of the journal, about which we knew nothing prior to publication.
The editors of the Stanford Law Review had told us that a couple of sentences in our very long paper had been changed after Lott’s reply was complete and asked us to restore the original language. We did so without objection. We understand from the Law Review that there was some additional word that Lott wanted changed that they did not ask us to change and that this disagreement led him to remove his name.
Since we were not aware of the dispute between Lott and the Stanford Law Review editors on this matter, we don’t know what the word might be, although it is hard to see how any single word could make much difference in a very lengthy article. But more importantly, when someone’s work is being identified as erroneous because of mis-coding errors, one would think that the focus of attention should be directed at either correcting the errors or showing that they do not exist. On this most crucial matter, we have heard nothing from Lott, and we are anxious to hear his response. This is particularly true because the same errors that are found in the paper from which he has removed his name also are found in his newly published book, and because he continues to lobby on behalf of concealed carry laws claiming that they will reduce crime, when his own regressions (when corrected) show that that is not true.
Teaser A number of readers raised what seems to me an important question: if the Ayres and Donohue statistical results showing a mild “more guns, more crime” effect were correct, what mechanism might account for that effect? Donohue and I have been back and forth by email on the question, and I’m hoping to have something ready to post later today or tomorrow.
Previous post here.