Read on Twitter
Some thoughts on the whole "ignore the polls" thing. I get that people shouldn& #39;t get complacent (please don& #39;t!) but I think there are layers here that are not part of the discourse. [1/n]
As someone who tries to follow the science, and believes in math, it& #39;s disheartening to have to constantly hear "the polls were wrong in 2016." There was a minor (typical) polling error, but they were pretty accurate. How the polls were *interpreted* by pundits was wrong. [2/n]
There is a difference between the polls being wrong vs. biased. Polls will be wrong, and errors can go either way (i.e. an error favoring Biden is as likely as an error favoring Trump). Of course, one matters more than the other: an error favoring Trump may change who wins. [3/n]
Polls being biased means that an error in one direction is more likely. There& #39;s no reason to believe that an average of polls (e.g. RCP, 538) is biased. So an error should be normally distributed. Polls can be wrong and the winner may not change. [4/n]
Trump is counting on the polls being biased, in more ways than one. Sure, he& #39;d like to win, but even if he loses he wants to be able to point to the polls and say they were wrong. After all, if both the results and the polls favor Biden, his case is weaker. [5/n]
Imagine Trump loses by 10 pts (I like to). To claim victory he needs to delegitimize votes. If there are polls showing Biden led by 10, it& #39;s harder to claim shenanigans. If people don& #39;t believe Biden actually led by 10, it& #39;s easier to claim that Trump& #39;s version is correct. [6/n]
So this means that "the polls are wrong!" actually helps Trump& #39;s post-election strategy. Reducing confidence in polls makes it easier for him to say that no one can know who voters actually preferred, other than the votes. And if not all the votes are counted . . . [7/n]
2nd point: comparing 2020 to 2016, showing that Clinton was a 70/80/90% favorite on a given day. First of all, an 80% favorite is not a lock. Second, Clinton& #39;s win probabilities were highly volatile: between 50 and 90% within the last month of the election. [8/n]
2020 looks more like 2012: Biden has been a steady favorite, his odds slowly climbing as he maintains his lead approaching election day. Not the same volatility, so not a fair comparison. Very much like Obama& #39;s lead over Romney, very consistent, leading to more certainty. [9/n]
Which brings me to my last point. In 2012, there was a lot of manufactured horse-race reporting, trying to make the election closer than it was because no one pays attention to the news if there& #39;s no conflict or fierce competition. [10/n]
I really really really hope that this is what is going on this year: not really close, but the press is making a big deal. I& #39;m not certain of that, though, because the GOP will do whatever they can, legal or not, to claim victory. [end]
