It’s pretty obvious to anybody who knows me, has read anything I’ve ever written, or heard anything I’ve said, that I’m hoping Biden will win the upcoming election. The polls all seem to predict that outcome, which is nice, but the polls predicted a solid win by Clinton in 2016, and that didn’t happen. That’s likely why some polls show a lead for Biden, but among the same sample, the prediction that Trump will win. How fascinating is that? Although a majority of those sampled prefer Biden, a majority believes Trump will win. To be fair, I might have been in the majority opinion on both questions, but I’m feeling more and more confident that the polls will predict the outcome. Here’s why.
First and foremost, I suspect that polling outfits learned a lot from 2016 and could see some of the mistakes that were made. Also, I’m still not convinced that there were all that many mistakes. The election was won by razor thin margins that were well within the error of most of the polls.
There were a few notable exceptions that made big differences. Wisconsin is one. The image here is a screenshot from Real Clear Politics showing the last few polls in Wisconsin before the 2016 election. Clinton was ahead in all of them by at least 6 points. In the end, Trump won by 0.7 points.
Plenty of people have blamed third party candidates (Johnson and Stein), and they did get enough votes to make a difference, but they got less than was predicted by the polls. Johnson’s RCP average was 5.8% and he ended up only getting 3.6%. Likewise, Stein’s RCP average was 2%, and she ended up with 1%. Yes, that 1% who voted for Stein would have tipped the state to Clinton if they all voted for Clinton, but Clinton was still polling ahead with more, not less support going to the third party candidates. There’s no real third party candidate on the ballot this year, so that’s very likely less of a factor.
But here’s where my comfort in the prediction from the polls comes, and something that’s important to look at when thinking about where any error might have been in 2016. Again, using Wisconsin because it’s an example of a state in which Clinton had a solid lead in the polls, and then lost. The polls predicted she would get 46.8% of the vote, and she got 46.5%. Pretty damn close. She was ahead by 6.5 points with 46.8% of the vote, and lost by 0.7 points with 46.5% of the vote. The error was almost entirely in the vote for Trump, which was predicted to be 40.3% and ended up being 47.2%.
Now let’s look at 2020.
The current RCP average has Biden up by 5.5 points, but with 49.5% of the vote, not 46.8%. If, like in 2016, the polls are missing Trump voters, but getting Biden voters pretty accurately, that’s going to be hard for Trump to make up the difference. Not impossible, because it could end up with Trump getting 50.5% and the polling average being perfectly correct for Biden (and also assumes that Biden’s numbers don’t improve as the race gets closer).
What about Michigan?
In 2016, the RCP average was a little better in Michigan (“better” meaning less error from the final outcome). Even though Clinton lost, the polls underestimated her support at 45.4%, and she ended up with 47.0%. But the error for Trump was greater, with him getting 47.3% in spite of a predicted 42.0%.
What about 2020?
Biden’s lead is stronger than Clinton’s was in the polls, and, again, he’s got an average of 49% to Trump’s 42.8%. Without a significant impact of third party candidates, in the polls and on the ballot, it’s certainly possible that support for Trump is going to be 51% and Biden’s numbers will be accurately reflected in the polls, but being so much closer to that 50% mark makes it notably different from 2020.
Even if Trump manages to win in Arizona and Florida and Ohio and North Carolina, if he loses Pennsylvania and Michigan and Wisconsin, he will very likely not have enough electoral votes to stay in office, and the map will look like the screenshot below.
This would mean 279 electoral votes for Biden (maybe slightly less if he loses all the votes from Nebraska and some of the votes from Maine), but still enough to win the election.
There’s a long way to go until November 3, and in typical 2020 fashion, so much can happen, but it’s seeming more and more likely that the predictions will be right this time.
We’ll know soon enough…or not soon enough…can it really get here soon enough????
One thought on “A hopeful read of some polls”
Pingback: Is the ride over or just getting started? – Hitting Bregma