Election 2020: Which Poll Came Closest?

Despite my long-standing reliance on Rasmussen Reports, I have decided to add two polls to my must-follow list: The Hill/HarrisX and IBD/TIPP. Rasmussen’s final poll before this year’s election had Biden leading Trump by 1 percentage point. In fact — assuming that the final vote count resembles the current tally — the nationwide count of popular votes puts Biden ahead of Trump by almost 4 percentage points. That was the spread predicted by The Hill/Harris and IBD/TIP in their final pre-election polls. If I had used that spread in my final projection, I would have nailed Trump’s share of the electoral vote, which now stands at 43 percent (before all results have been certified and all court challenges have been heard and ruled upon).

In fact, this year’s (apparent) result is exactly in line with the equation that I had derived from the results for the elections of 1972 – 2016:

With the addition of 2020, the relationship between popular-vote share and electoral-vote share looks like this:

The only change is a slight improvement in explanatory power (r-squared rose from 0.92 to 0.93).

The GOP continues to hold an edge in the electoral college, but it is a slight edge. According to the equations in the graphs, a GOP candidate must muster at least 49.5 percent of the two-party popular vote to be sure of winning an electoral-vote majority.

Trump got lucky in 2016. Because of razor-thin victories in a few key States, he got 56.9 percent of the electoral vote with only 48.9 percent of the two-party popular vote (i.e., a deficit of just over 2 percentage points).

This year, however, Trump seems to have eked out only 48.1 percent of the two-party popular vote (i.e., a deficit of almost 4 percentage points), and the close calls (apparently) went to Biden. Result: A reversal of the 2016 outcome.

For more about the accuracy of various polls, see this piece at NewsMax (behind a paywall). Here’s some of it:

The Investor’s Business Daily/TIPP poll defended its title as the most accurate pollster for predicting presidential outcomes. The pollsters take the No. 1 spot for the fifth presidential cycle in a row, a Newsmax review reveals. Among the worst polls were those from CNN and Quinnipiac.

One of only two polls to predict President Donald Trump’s 2016 win, the IBD/TIPP poll came closest to predicting the outcome of the 2020 presidential election. The Hill-HarrisX poll also predicted election results with the same accuracy, according to American Research Group, Inc….

Also making the top of the list were Emerson and Rasmussen Reports — one of Trump’s favorite pollsters. Fox News, USA Today/Suffolk University, and New York Times/Siena College came in at the middle of the pack. Among the worst polls were Economist/YouGov, CNBC/Change, NBC News/ Wall Street Journal, USC Dornsife, Quinnipiac, and CNN. All predicted Biden would lead Trump by double digits.

In any event, among the many sources that I will never consult is Nate Silver’s overrated statistical mishmash called FiveThirtyEight. Silver predicted not only Democrat gains in the House (wrong) and a likely flip of the Senate (probably wrong), but also a 9-point spread between Biden and Trump in the nationwide tally of popular votes (very wrong).

Election 2020: State of Play and Accuracy of Polls

UPDATED AT 4:00 PM CST, 11/04/20

In yesterday’s post, I forecast a 285-253 electoral-vote victory for Biden. As of this morning afternoon, according to FoxNews, Biden has locked up 264 electoral votes and is leading in three States (Michigan, Nevada, and Wisconsin) whose electoral votes would give him a total of 270 — just enough for victory. It’s possible that Biden’s total could be higher, given the number of votes still to be counted in Georgia, North Carolina, and Pennsylvania, where Trump is currently ahead. And, of course, Biden could still lose any or all of the three undecided States where he is currently ahead. And both candidates can be expected to demand recounts, and recounts of recounts, and to seek judicial intervention all the way to the Supreme Court. But, whatever the outcome, I am pleased by the accuracy of my forecast. (Though I will be most displeased by the outcome if Biden proves to be the winner.)

I based my forecast on polling conducted by Rasmussen Reports, which has been my go-to source for presidential polling for the past 12 years. Rasmussen has acquired a bum rap for being pro-Republican because its generally accurate polls aren’t biased toward Democrats as are most other polls. This year’s presidential race provides further evidence of Rasmussen’s lack of bias.

As of now, Biden has a 2 percentage-point lead over Trump in the nationwide tally of popular votes. That lead might rise a bit when all of the Left Coast votes have been counted, but it’s statistically the same as the 1-percent edge forecast in Rasmussen’s final poll. And how did Rasmussen do relative to other major polls? See for yourself: