Read on Twitter
1/The "study" is without merit. It would receive a failing grade were it submitted as research paper for a graduate course in any decent doctoral program in economics. (Certainly the same is true for the political scientists and sociologists with whom I interact.)
2/By conditioning on a county where Trump did poorly/fraud was alleged, the paper in essence interprets the unpredictable part of that poor performance as fraud. This is conditioning on unobserved heterogeneity that affects Trump support and raises classic selection problems.
3/Lott's examination of differences between in person and absentee voting is supposed to address this. But his strategy amounts to what I have called "identification by lack of imagination."
4/From the paper, "While Democrats were pushing their voters to vote by absentee ballot, there is no
reason to expect that rate to differ between two precincts that are next to each other and are similar in terms of their in-person voting support and their demographics."
reason to expect that rate to differ between two precincts that are next to each other and are similar in terms of their in-person voting support and their demographics."
5/Why? In an election where absentee versus in person voting has itself been politicized, it is obvious that rates can differ because of differences in opinions of the politicizer in chief, i.e. Trump. In econospeak, the paper does not have a theory of voting modes or choices.
In econospeak, the paper does not contain a theory of voting modes or choices. It in essence makes up instruments/controls based on hunches.
7/As for the 2016 and 2020 comparisons as another form of control, these are useless. Basic exchangeability requirements are violated. 2020 had 4 years of voter exposure to Trump and a cohort of young voters, where Trump support exceptionally low.
8/I could go on, but honestly I abhor the use of twitter as a forum to criticize papers. I do so now because of the grotesque irresponsibility of releasing the paper under guise of scholarship.
9/ Let me end with the following. Consider a paper that studies Black white wage differences across occupational categories with similar skill requirements.
10/Suppose its strategy for identifying discrimination was to take the category with the largest BW gap and compare it to others, albeit with crude controls of the type found here, and used the gap wrt other categories as a measure of discrimination.
11/You can be sure that scholars (such as myself) who believe that discrimination is a core component of racial inequality would be critical of the. Would those taking the Lott study seriously be equally credulous if the context were racial inequality?
