Wednesday, October 21, 2020

13 Days and Counting

 Joe Biden's lead in the RCP polling average dropped to 8.6% yesterday, as a result of an Investors's Business Daily poll that put his lead at only 3%. Some of the races in the battleground states have also narrowed, so Biden's average lead in those states is now just 4%.Meanwhile, as of this morning 38 million early votes have already been submitted.

In yesterday's post I mentioned the polling phenomenon called social desirability bias, where people answer poll questions the way they think others will approve, not necessarily the way they will actually vote in the end. A polling outfit called Trafalgar Group tries hard to overcome this bias in various ways. Trafalgar is one of the very few that got 2016 right (exactly predicting the Electoral College outcome). 

This year Trafalgar is predicting an EC total for Trump in the 270s. That would be enough to win, but it would be a very close call. In such a scenario the results in just about every close-run state would be contested in the courts, because any one state flipping the other way would alter the outcome of the race. 

I'm not presuming Trafalgar is right this year, mind you. They only have to be wrong in one state that they pick for Trump in order to reverse the resulting outcome. But the real takeaway is that this race, according to Trafalgar, is much closer than virtually all the other pollsters are suggesting.

How the Polls Hide Trump's Lead discusses how this could be so, looking at not only Trafalgar but at a British polling outfit called Democracy Institute, which also called 2016. This year they're saying that Trump leads in virtually every battleground state!

These pollsters suggest that most of the other polling organizations have not done enough to overcome social desirability bias, which is especially pronounced when the divisions are along class lines. I'll just say that if they're right, this may turn out to be a kind of Waterloo for the the whole concept of polling as a reliable means of monitoring the public mood.

* * *

Read also:

Liberal Anxiety Over a 2016 Repeat May Be Why We Won't Have One

Ten Signs Trump will Win Again


No comments: