Editor’s Note: As discussed last week, I am going to be sending more frequent updates about the campaign as the 2020 cycle comes to a close (or to a climax?). If you have friends that you think would benefit from this newsletter, please do forward it to them.
Election-watchers largely agree that the end of the presidential campaign begins on Labor Day. The fall portion of the contest is certainly the most covered by the media. The final months of the cycle have also tended to be volatile historically, with lots of noteworthy (though rarely, if ever, decisive) events such as debates and October surprises.
Yet the hum-drum of the fall election campaign obscures a simpler truth about presidential contests: The vast majority of voters have made up their minds by now, and campaign events only tend to cause final outcomes to deviate slightly, if at all, from the polls today. Since 1952, no candidate has ever won the popular vote after trailing on Labor Day. It should be noted, of course, George W Bush and Donald Trump nevertheless won the electoral college majority with a minority of the popular vote, though both of those elections were already close at this point in time.
The passage of Labor Day this Monday thus means that it behoves us to start paying closer attention to polling data and other election indicators. Let’s start off this first brief of the final weeks of the 2020 campaign by looking at polls and election forecasts. We’ll talk in more detail about other electoral indicators and the feelings of the electorate toward the issues of the day in my next missive on the 2020 campaign.
According to the polling averages from FiveThirtyEight and The Economist, Joe Biden leads Donald Trump nationwide by between 7 and 8 percentage points. The simple election model I put together for educational purposes currently gives Trump an eight-and-a-half percentage point deficit:

Needless to say, he is very unlikely to make up that ground. Polls this year have been historically stable. Even amid civil unrest, the coronavirus pandemic and economic collapse, Joe Biden’s projected vote margin has moved with a standard deviation (a measure of the “spread” of data) of just 1 percentage point. The volatility of the average election cycle before 2020 was closer to a 3 or 4 point standard deviation.
This leaves Trump mainly with the prospect of winning a majority of votes in the electoral college with a minority of the popular vote — a repeat of his minoritarian victory from 2016.
This means we should focus on the key states to his victory. If we rank the states by their likelihood of providing Trump or Biden with their 270th electoral vote, we should pay extra attention to Michigan, Nevada, Wisconsin, Arizona, Pennsylvania, Florida, and Texas. Again according to my toy model, here is a rough idea of the state of public opinion — and the margin of error — in these states, plus some others:

Though the race varies in precise magnitude by state, Joe Biden has a clear lead in all of the main key states (FL, PA, AZ, WI, MI, NV and MN) and is tied or just barely behind in the others (NC, TX, GA, IA, and OH). This shows up in both the FiveThirtyEight and Economist models too:


Adding these up, Biden is currently favored to win in states worth 334 electoral votes. If the election were held today, he would almost certainly win.
Of course, the election is not being held today. There are 55 days until the polls close. How much can happen by then?
Looking at the following plot (source) of error in election polls by year might help gauge expectations.

First, note that modern elections have tended to be less volatile and more predictable at this point in the race. At 55 days out, the average error in polls since 2000 is just 3 percentage points on margin. Since 1980, it’s closer to 5. An error of the former magnitude would be enough to make the election close but steal leaning toward Biden. A 5-point shift toward Trump would put us in toss-up territory.
It is here that I leave you with a question. How volatile do you think this election will be? Based on the polls over the past 6 months, I’m guessing not very. Accordingly, the models I have built for the election are looking good for the Democrats. Again, here’s the toy model:

To put that in context, that at this point in 2016, the same toy model would have given Donald Trump about twice the chance.
So, do you think that the election is relatively set in stone already? Or do you think we’re looking at a shift in the polls as large as the type we saw in the 70s and 80s, when there were twice as many swing voters in the electorate? Let me know your thoughts in the discussion section below (and don’t be afraid to engage with other readers!).
Until next time,
Elliott
Your snake model is wonderfully vivid. Nice work.
I want to see what the race looks like in a week. If Biden regains some of his pre-convention support because Trump's convention bounce is fading, Trump has major problems. Trump's convention bounce didn't have Trump take the lead in keys states like PA and FL. If this is supposed to be one of Trump's best polling week or best two weeks, he isn't in good shape.
There could be an September/October surprise which could change of course of the election. The Woodward Book? Ron Johnson's committee investigation? The Durham Report? Or maybe the debates change the election? Due to polarization, maybe these potential events won't change the election.
My guess is the election won't really move that much for the rest of the race and Biden is headed for a 7 to 8 percent popular vote victory and will flip MI, WI, PA, and FL, and likely AZ and perhaps NC. We'll see.