Whether you are dooming, blooming, or somewhere in between, if you follow the US election, you are noticing the poll numbers.
It’s election season, which means one thing: The polls are out in force. New ones are being released every day, sometimes several of them. There is no shortage of data and between now and November 5, it will continue at pace or increase.
However, there is also a prevailing notion in some camps that people need to hang on each poll’s findings and ride the ups and downs of the margins. This is not a great idea for your mental health, nor should you from an analytical perspective.
Poll Numbers: Reality Check
In recent times, I have become familiar with the work of Carl Allen, an analyst who has written a new book about examining and understanding poll findings. He has a number of good insights, and two of them are as such: (1) Individual polls are not to be taken as exact predictions of margin or outcome, and (2) One poll is not necessarily indicative of anything. Allen had a video from September 2024 in which he explained, using a simulated environment, how you could get some widely varied individual results but they all average out to a tie, with a handful of undecideds that hold the key.
To that end, let’s focus on the Quinnipiac poll – and yes, we are singling them out for this explanation, as that is what prompted this article. In their September 12-16 poll, they had Kamala Harris ahead in Michigan by a margin of 51 to 46 with three percent undecided. About two weeks later, in their October 3-7 poll, they had Donald Trump ahead 51 to 47 percent with two percent undecided. Do you think there was actually a four or five percent swing of decided voters in the span of two weeks, with no major intervening political event or “October Surprise?”
What if Quinnipiac’s next poll has Harris back over 50 percent in Michigan, or the race tied – did those voters “switch back?” Does a not-insignificant segment of the likely, decided electorate flop in the wind and change their minds weekly?
The answer to all of the above is probably not. Sure, Donald Trump could win Michigan, as could Kamala Harris. It’s close. However, one cannot ride the wave and hang on every single poll like it is the gospel truth. If you are hoping one particular candidate wins a race, that’s a good way to drive yourself insane. One bad poll and you’re low, followed by one good poll and you’re high again. It’s the summary of all the information averaged together that is more important, as well as allocating how the undecideds are likely to break based on demographic information available in the individual crosstabs.
Pollster Quality
Keep in mind also that just like in 2022, there is a wave of partisan and/or relatively unknown pollsters churning out results. Based on those surveys, for example, RealClear Politics rated the 2022 Washington Senate race as a toss-up. A bunch of GOP pollsters had that race close in the final stretch, with Democrat Patty Murray under 50 percent, but Murray ended up winning by almost 15 points. This also ignored the fact that she cleared 50 percent in the open primary and never should have been considered in serious danger. (In fairness, most other prognosticators had the race as Likely Democratic.)
This is, however, an example of why the quality of polls should be taken into consideration as well.
Don’t Sweat The Individual Poll Numbers, But…
If an individual poll’s findings seem too good to be true, they might be. If the crosstabs of an individual survey seem illogical or outside expected norms, compare it to other reputable polls. Look at the big picture. Consider sample size, margins of error, the biases of the pollster, and so forth.
This is a lot of words to say: Don’t freak out about one poll. I know you are; I see you over there freaking out. Yes, you. Take a breath, fire up the Bartesian machine, and watch a movie.
