The news media is fully focused on the Georgia runoff election for two Senate seats. This election coverage has one major difference from any such coverage of the last 75 years. There is almost no mention of polling data, except to occasionally say “it’s within the margin of error”. Why would that be? The polling industry had one last chance to redeem itself in last November’s election. It failed. Political media does not want any more egg on its face created by glossy graphic displays of polls that turn out to be wrong.
Due to a huge plurality for Biden in California, the polls almost got the popular vote margin correct in the 2020 presidential election. They missed almost every other competitive race. The battleground state polls, the only ones that mattered, were wrong. Biden won most of the races, but not by anywhere near the predicted margin. The track record was much worse in the legislative seat races. In some cases, the polls were off by double digits. I don’t remember one poll predicting a win for Susan Collins. She won in a landslide, despite being outspent 2 to 1. I am unaware of any pollster who predicted a 10 seat pickup in the House for Republicans. Results in the race for state legislatures was even worse. The Democrats were supposed to flip a number of states; they flipped none and lost one. What’s going on?
My first hint came in 2016. Several people I know said they were going to vote for Trump, but I was sworn to secrecy, because they were afraid their neighbors would never speak to them again or their careers would be adversely impacted if their preference was revealed. Eventually, this phenomenon seeped into the consciousness of the media and was labeled “shy Trump voters”.
The second problem is the polls themselves. The pollsters are social scientists. Social science isn’t akin to physical science. Human behavior does not lend itself to a fully controlled experiment, and such an experiment would be illegal and/or unachievable in any but a totalitarian society. But the practitioners like to play at science. The polls are too long and wallow in unproductive detail in order to seem scientific. I hang up on most polls after 5 minutes. I warn the caller that I will only devote 5 minutes and they plunge, per instructions, into a 20 minute poll. The difference between very, somewhat, slightly, not at all and REALLY not at all does not add to the sum of human knowledge. No shortcut answers are allowed.
The third problem is that a great many voters believe that polling organizations are part of the media/entertainment/intellectual arm of the Democratic party. They think the published poll results are designed to motivate Democratic donors and voters and discourage Republican donors and voters. Given the fact that almost all of the egregious mistakes made by the polling industry predicted wins by the left side of the political spectrum, that suspicion has some justification. If you think that the polling organization calling on you is part of the righteous gaggle that regards you with disdain, and is attempting to discourage you from going to the polls, why would you cooperate? Disdain is a powerful force that produces and equal and opposite reaction. Many of those polled refuse to answer or lie to the caller.
The polling industry will either need to come up with a new model or fold its tent. Its disappearance might be a good thing. Politicians spend way too much time consulting polls and consultants who parse the polls. More time listening to constituents might be a useful upgrade in terms of time management and mindshare.