Read on Twitter
The DC rioters said they were trying to “stop the steal”. Their views were stoked by various claims from the Trump campaign, some of which are obviously false. But @JohnRLottJr's paper looks like social science. So @grimmer, Haritz Garro, and I decided to engage with it. 1/n
Lott posted his paper in December. Trump and others promoted it. It was touted by @realPeterNavarro as a game-changer.
On Tuesday, we posted a critique of the first part of the paper. https://twitter.com/justingrimmer/status/1346144084902072320
The next day @JohnRLottJr acknowledged we were right and said he would retract that part of the paper. But he claimed the second part still holds. It doesn’t.
The second part of his paper looks at county-level turnout in 9 states. It purports to show that turnout was inexplicably high in 19 counties where, after the election, Republicans claimed that fraud had taken place.
Lott’s conclusion would be reckless even if his finding were robust. He finds 1-2 pct “unexplained” turnout in these counties. Could be GOTV efforts; or maybe Trump’s team focused their allegations on counties with high turnout. But Lott says it’s fraud.
Now, if you look at the raw turnout data by county for these states, there doesn’t seem to be anything out of the ordinary. Turnout in 2020 is mostly *lower* than you would expect in the "suspect" counties (shown in red) relative to 2016 turnout.
To get Lott’s finding, you have to include three other states (OH, NC, and FL) where turnout was lower in 2020 relative to 2016. Turnout appears higher than expected in the “suspect” counties only because Lott compares them to counties in other states where turnout was lower.
Lott’s result (and thus his evidence for fraud) goes away if you exclude those three states or account for different turnout trends across states.
We show that Lott's analyses are biased toward finding excess turnout in these counties: if you chose counties at random rather than use the ones Lott singles out, you would reject the null around 70% of the time and on avg you would get a larger t-statistic than he does.
Refuting Lott’s findings is obviously not going to change Trump’s mind. And maybe it wouldn’t have changed anyone’s mind who stormed the Capitol.
But given how many Americans appear to doubt the integrity of the recent election, we think it’s important to set the record straight.
Our updated paper: https://www.dropbox.com/s/ps5oufevtka3nmf/Fraud.pdf?dl=0 /end
Our updated paper: https://www.dropbox.com/s/ps5oufevtka3nmf/Fraud.pdf?dl=0 /end
Wrong tag -- should be @justingrimmer.
