In 2016 I used a model to make the following prediction:
Trump odds of winning: 71%, with an expected electoral vote total of 301. He won with 304 EV and the rest is history. Nobody else came anywhere close to being this accurate, not even High Nate Silver.
The model worked like this: I took the latest available poll averages for each state from FiveThirtyEight.com, then did a series of trials where I added between 0 and 5 percentage points to Trump's poll results to determine who won each state's electoral votes. In the case of a close call (within a percentage point) I split the EV equally between the two. That reduced the impact of random factors like the weather on very tight races.
The 0-5 point boost for Trump was based on documented differences between phone and online polls, also known as the "shy Trump voter" effect. The FiveThirtyEight polls were primarily phone polls.
Now that we're less than a month away from the election, what does the model say? Using the exact same methodology as in 2016, the prediction is this:
Biden odds of winning: 65% with expected EV total of 291. [[POSTSCRIPT: He got 306. Close enough?]]
The poll data are mostly from the last week of September, shortly after the debate. So, the prediction can be viewed as the likely outcome had the election been on say October 1. I might update this later on or I might not. It's a lot of work transferring the poll data into my spreadsheet.
A lot can happen in a month (from what I recollect, last time the model was showing a modest Clinton advantage until almost the final update.) But this year is different. Really different. I expect that a candidate will need at least 320 EV to avoid a disputed result that, if we are lucky, will go like it did in 1876, and if we are not lucky, well, I don't even want to think about it. The model shows a 77% likelihood that neither candidate will reach this threshold.
I earlier predicted a 60% chance that we don't know who the president is on 1/20/21. I'm sticking with that, assuming that 77 times out of 100 (I am a frequentist on Tuesdays) we get a dispute but that only 60 times out of 100 that dispute is unresolved by Inauguration Day.
Here I'll repeat my recommendation not to vote. Your vote isn't going to be decisive. I don't want to hear about how some dogcatcher in Kalamazoo lost by one vote because he forgot to vote for himself - your vote is not going to make a difference. Given that it isn't going to be decisive, the only effect of voting will be to mentally invest you in one of the candidates and degrade your ability to act in your own interest after the election.
