Reader Question: Does Our Model Factor In Rejected Ballots?
A.S. from New Jersey: Have you considered factors such as rejected absentee ballots (which would affect Democrats more because they’re more likely to use those) in your probability estimations?
No, our model doesn’t attempt to account for anything that happens after ballots are cast, such as mail ballots being rejected or Trump successfully challenging the outcome of the election. We’re not trying to dismiss these concerns; they’re just outside the scope of a statistical analysis. Things like human error or the whims of the president can’t be modeled!
If you’re interested in reading more about rejected ballots, I humbly recommend this article I wrote last week. It’s certainly concerning for democracy that hundreds of thousands of otherwise eligible voters will probably see their ballots invalidated this year; however, I would be surprised if it actually affected the outcome of the election. First of all, only 30-40 percent of ballots are expected to be cast by mail this year. Second, only about 1 percent of mail ballots are rejected. That means an election would have to be within 0.3 or 0.4 percentage points for rejected mail ballots to be decisive. And even in that case, it’s not as if all mail ballots are cast for Democrats. So even a Democratic lead of 0.3 or 0.4 points probably wouldn’t be erased (just narrowed) by rejected mail ballots.
Of course, this is a very rough, hypothetical example. In some states there will be a lot more than 30 percent of votes cast by mail. Furthermore, the rejection rate could be higher this year because so many people are voting by mail for the first time and are unfamiliar with the rules (although experts are optimistic that this will be mitigated by states extending their ballot-receipt deadlines, allowing voters to fix mistakes on their ballot and making ballot instructions clearer). So we can’t totally rule out rejected ballots being decisive in a hyper-close race. I just think it’s unlikely.