It Looks Like Another Troubling Year for Polling
November 6, 2022
Nate Cohn has some important words of caution for us about the Republican gaming of the polling averages, which he has already described as something that is real despite some nonsensical and defensive tweets from the other Nate (Silver) at FiveThirtyEight. Silver has come under criticism for ignoring what’s happening, suggesting that Democrats should just do the same thing – a comment that should make people wonder if they should ever take this guy seriously again. Another thing Silver has said is that since the Democrats have not done the same thing, this means they have internal polling that is bad for them and are not releasing it. This makes no sense because the Republican polls that are gaming the averages are not internal Republican polls but from GOP-aligned polling outfits. Silver’s bizarre and combative defensiveness suggests that he knows he has a problem with FiveThirtyEight’s polling averages. The one thing that Silver’s comments tell us is that Republicans are gaming the averages.1
But back to the first Nate (Cohn). In the latest issue of The Tilt, Cohn suggests that rather than hiding some Democratic advantage, the Republican gaming of the averages might actually balance it out. Why? Because if the polling still contains the errors we saw in 2016 and 2020 (but not in 2018, when the polls were fairly accurate), then the reputable pollsters are still making errors that are biased in the Democrat’s favor. That’s not intentional on their part, as the error in the Republican polls certainly appears to be. (If these GOP pollsters are not making stuff up and sometimes getting lucky, then they can be transparent and release their methods, toplines, and crosstabs like reputable pollsters do. Until they do that, they are making it up as far as I am concerned and where they hit they are just getting lucky.) If reputable polls have been biased towards Democrats – seemingly to the tune of 2-4 points – it is because they are not reaching enough Republican voters. The Republican pollsters seem to just add points in the Republicans favors because they think that’s what’s true. Sometimes they are right, but sometimes they are wrong. It starting to become part of the media narrative that these GOP pollsters are always right in their hunches, but the data in individual races shows that’s not true.
There is a problem reaching Republican voters for surveys. That’s well documented at this point. There is no evidence that respondents are lying to pollsters in any significant way (i.e., that would impact the results), so we can put that argument to rest. The problem is Republican voters refusing to participate in polls, which creates problematic weighting challenges. But another problem is electorate modeling. And this is how pollsters may be underestimating the Democratic vote this year. Even the reputable pollsters have been modeling electorates that are more Republican than the early voting data and recent voter registration data suggests the actual electorate might be. To be fair, the early vote is older than it was in 2018, but it appears that a lack of young Republicans is the reason for it. And Democrats turned out a higher rate of seniors in the early vote than did Republicans. In any case, young Democratic voters will need to turn out on Tuesday for Democrats to win. It could be that the Election Day electorate will be so Republican that it won’t matter. Only time will tell if that’s true. In the meantime, it is a problem for pollsters.
So, what’s this mean for us right now? As I wrote a few days ago, we have to ignore the polls and focus on mobilizing voters to get out and vote. And the media should be preparing the public that Democrats could have a good night on Tuesday since it is possible – and there is evidence to support it – and it seems clear that Republicans are trying to convince voters that there is no way they can lose so if they do, guess what? The election was stolen! We already have enough danger to democracy without radicalizing more new fascists to engage in violence. While the blame for any violence falls squarely on Republican leaders who are egging it on, the media has thus far failed miserably in its role as a guardian of democracy.
In many places, the early vote is over or ending today. Election Day is Tuesday! Mobilize and Vote!
The question has come up of what purpose would it serve to game the polling averages. I’ve noted previously that it can lead to depressing Democrats who give up and don’t work hard to mobilize voters and maybe even don’t vote. Dan Pfeiffer has a really good insight into why: “[C]reating a false sense of momentum is a long-running Republican strategy. Many Republicans believe the best way to win elections is to convince voters that you are already winning. The strategy is loosely based on the idea of social proof – people want to be with the winners, not the losers.”