Crow Eating Post
Although I am immensely relieved at Joe Biden winning the Presidency, I am here to admit that the polls flat sucked. No other way to put it.
I defended polling after 2016... and I generally DO believe it got a bad rap. Polling in 2016, besides in a few key states, was not bad. Polling in 2018 was pretty good. I was confident that 2020 polling would be even better. I was very wrong.
The final RCP average had Biden up by 7.2 points. He won by 3.1. This is a big, fat miss. What made it even worse, though, was the state level misses were absolutely inexcusable. How did Susan Collins not lead in a single poll all election cycle and end up with a 10 point win in Maine? I get that there may be "shy Trump voters" (a phenomenon that I had dismissed, but now will have to take very seriously)... but there are "shy Susan Collins voters"? Really? And the FINAL kick in the nuts was that the polls were all off in the same direction. In 2016, Hillary actually outperformed her polls in some places, such as Arizona and Texas. In 2020, polls underestimated Trump support pretty much across the board.
My Grades:
The Polls: D
After all the bad things I said about the polls, I can't give them an F. Why? Although they were OFF BIG in terms of how close the race was going to be, they actually told the correct narrative; Joe Biden was likely to win the Presidency by taking back the industrial Midwest. This is exactly what happened. Further, they told us that he had expanded the map and was playing offense in some Trump states. Polls said he had a shot at flipping North Carolina, Arizona, Georgia, Texas, and Florida. This was true in three out of those five places. So, a very bad day for the polls, but not quite an
unmitigated disaster.
Nate Silver/ fivethirtyeight: C-
Nate Silver's model is only as good as the information that gets fed into it. That being said, just like the polls, he told the right story overall. He said that Biden was ahead of Hillary's pace and that there would have to be an even greater polling error than 2016 in order for Trump to have a shot, but he would still be an underdog. It turns out that there
was an even greater polling error than 2016, and Trump
did have a shot, but he
was still and underdog and lost.
Further, there is some posthumous vindication for Silver's 2016 call. Silver gave Trump approximately a 30% chance of winning in 2016. Biden ended up outperforming Hillary by just
one point in the popular vote. That one point, however, Translated into winning FIVE more states. In other words, it shows just how statistically unlikely Trump's win in 2016
was.
My beef with Nate Silver is that he is going to really have to rethink how he grades and weights pollsters. Many of the pollsters that he gave an A grade to for their methods had the
worst misses. Other pollsters who he did not include because he deemed their methods unscientific (Trafalgar) were actually more accurate than the "highly scientific" gold standard pollsters. What does this say about the state of polling in 2020? Well, I have several theories, but this post is already too long, so I will save them and just say, in a word: it's a mess.
In conclusion, there is an old question that comes up every once in a while in the WR: have you ever changed your mind on anything related to politics due to new information? I can say, yes. 2020 changed the confidence I will have in polls drastically. 2016 really didn't... and I don't think it
should have. But 2020 is a different story.
All that being said, I still think I like this election better than
@Starman