The first of Mr. O’Connor’s Three Rules of Science™ is:
all measurements are uncertain.
It’s the size of the uncertainty that matters. How much uncertainty can you live with?
Some phenomena can be measured quite precisely. Others, not so much.
Measuring people’s feelings, for example, or their intentions, is difficult. Even when faced with a binary choice like Trump v. Biden people equivocate when asked about their decisions.
This creates uncertainties in polling. We all KNOW there are such uncertainties but we don’t talk about them. We look at a poll or a survey and it says “47% of people support . . .” and we don’t think “oh that means it could be 42% or even 52%” or somesuch. We get stuck on the number as a fixed thing.
It’s not. It’s just a stop on a continuum. The polls all have ranges attached to their numbers like “+/- 3%” and that means that’s the most likely set of outcomes in their model.
Polling and pollsters are going to get a lot of heat in this election, much like the last one, but it is misplaced. Consumers of polling information should focus on the uncertainties in the polling results and the biases in the polling methods and not assume the models are predictive. They are just models after all, and even good models need to be continuously tweaked.
A “surprise” in an election is often just dissonance with the polls. An expectation of an election result is formed by the pre-election information presented by the polls. If you did not have that information in the first place you might not consider the outcome a surprise!
We expect a lot from these polls. We expect them to give us knowledge about the future when they can only guess at a cluster of possibilities. It’s the process of making those predictions that’s exciting, not the predictions themselves. Building robust, powerful models is foundational work in science. That can only be done with continuous trial-and-error. If the pollsters get it “wrong” then they have a new challenge to work on for the next go-round.
It stinks to have this much uncertainty in our national election. It’s hard to live with. What matters at this point, of course, is not what anyone said beforehand, but reducing, as much as possible, the uncertainly in the final vote counts. Even something as apparently simple as counting and tallying has uncertainty, and I’m not sure any of us knows how much of that there is, and I’m also not sure we really want to find out.