It’s not the polls, it’s how voting is skewed by the polls.
I suspect part of the problem with the voting not following the polling is the polling itself. I’ll explain with a sports analogy.
If a team is down but within reach, they are motivated to try really hard to catch up. And they frequently do. Only to relax, and immediately fall behind again. If a team has a seemingly large lead, they tend to become complacent and play conservatively, allowing the more motivated team that is behind to catch up. Once the team that had the lead are behind, their complacency disappears and they are motivated to try harder. The change in lead changes the motivation of the teams, and with it the momentum in the game.
What this implies in relation to this article and polling is that what potential democratic voters took away from the polls was the possibility that it could be a landslide victory, and allowed their optimism to flip a possibility into a certainty in their minds. Some of these voters relaxed and became complacent, and may have decided not to vote since hey, victory was assured. Meanwhile Republican voters looked at the same polls, saw that there was the possibility of a close race, and their brains switched a possibility of a close election into a certainty of a close election, motivating them to vote since in a close race they could make the difference. In other words, because the polls told them they were close behind, Republican voters were motivated to vote. The same polls told Democratic voters they were well ahead, leading to complacency, so some didn’t vote.
The end result is even if the polls were accurate in tapping into people’s intentions to vote, the predictions made on the basis of those intentions fed-back into the decision to vote, motivating some Republicans and demotivating some Democrats. Even if only a percent or two of each, that is more than enough to swing tight races one way or another.
A lesson is that polls aren’t independent, they interact with the public's perceptions of what is happening in an election campaign, which influences their motivation and therefore their decision to vote. The end result is motivating voters for candidates slightly behind in the polls, and demotivating voters for candidates ahead in the polls, resulting in more upset victories and closer elections than the polls would have you believe. Incidentally, this is a scientifically testable hypothesis.
This implies the problem stated in the article, that “The real catastrophe is that the failure of the polls leaves Americans with no reliable way to understand what we as a people think outside of elections — which in turn threatens our ability to make choices, or to cohere as a nation” may not be the problem. The problem may be that the polls are a reflection of what people think, but the polls themselves skew voting enough that voting is no longer a reflection of what the people think. The problem is that the polls make predictions, and the prediction result in differential feedback to the voting public, motivating the minority faction and demotivating the majority faction. The team that’s behind tries harder, and the team that’s ahead take their foot off the gas.
