Why You Should Really, Really Ignore the Polls

The final days of a presidential campaign can be stressful. “Will there be an October Surprise?”, “What will the undecideds do?”, “Will I have to move to Greece?”. So many things to worry about.

There is stress every election. But it’s somehow worse this year when we’re ruminating less on what our tax rate will be next year and more on whether we’re all going back to camp, and I don’t mean the sort of camp I used to go to where I taught cute Jewish girls how to play tennis. The camps Trump seems keen to send us to probably won’t have names like Pasaquaney or Rustic Adventure. His camps will be more like “The Stephen Miller Center for Reeducation”, although, oddly, they’ll still be big on lanyard making.

Perhaps the greatest source of pre-election angst are the polls, which is unfortunate, because at this stage of the election, polls are ubiquitous. There are national polls, swing-state polls, polls of different demographic groups and, of course, the dreaded aggregations of polls.

New polls are released every couple of hours and they can give your brain the feeling of a ping-pong ball. Harris will be up by 4 points in Michigan, and you’ll go out to grab yourself a celebratory beer. And then you will glance at your phone and see another poll has her down 3 in Michigan, and you’ll spill the beer all over your lap, making you look like you peed your pants. And then you’ll actually pee your pants, making you grateful that you spilled the beer to hide it.

A national poll will show Kamala up 5, her largest lead of the race, quickly followed by one showing her down 2 as Steve Kornacki helpfully tells us that Trump is “surging among in Inuit and Yupik people on the Kenai Peninsula”.

This is, to be clear, crazy-making. But there are a few things about polls which you should know. And if you know them, it might help you either tune them out, which is generally the correct course of action, or at least keep them in perspective.

First, polls suck, and in many ways, they are getting suckier.

In order to conduct a poll, the pollster has to take a number of guesses. Let’s say that we know that right-handed and left-handed people tend to vote differently (God, being politically correct is so freakin’ hard these days). And let’s say that right handed people, who are on average smarter (I am right-handed by the way) make up 65% of the population, while left-handers make up 25% and ambidextrous folks 10%.

You probably understand that your poll should be made up of 65% righties, 25% lefties, and 10% bothies to have an accurate representation. BUT, that may well be wrong. Because let’s say that righties vote 10% more frequently, on average than lefties (on account of being smarter and all). In that case, righties should make up about 70% of your survey. But of course, the recent averages that you are basing this on may be wrong, because there is a left-hander on the ballot this year who might draw more dim and simple left-handers to the voting booth. So how will you account for that?

In other words, on this variable and dozens of others, pollsters have to literally just make a somewhat educated guess as to what the electorate will actually be. And each pollster makes a different guesses.

Will the Dobbs decision result in a higher than normal turnout among women voters? If so, how much higher? And will those extra women be college educated? Will they be black? White? Are there geographic differences? Will these newly energized women be right-handers who can find the ballot box, or left-handers who find themselves lost in random parking lots, unable to read even the simplest signs?

Again, there are dozens and dozens of variables about which the pollsters must make guesses (they call them assumptions), so one of the big reasons one poll’s result differs so much from another’s is that they’ve just made a lot of different assumptions. Who is right? We won’t know until election day. But in recent elections, a lot of pollsters have made a lot of bad assumptions.

Take the Trump-Hillary election. Hillary was winning by a lot in most polls heading into election day. But she lost. Why? Because historically not-college-educated rural voters turned out in low numbers.

But something about “I’m so rich” Trump and his golden toilet resonated in economically distressed rural America. Whereas most pollsters assumed they would comprise (I’m making up the numbers) 15% of the electorate in 2016, they actually made up about 20%, dramatically changing the results. The polls were right about how various groups planned to vote. They simply made a wrong assumption about the relative size of these groups.

The same thing happened in 2022 when the predicted “red wave” never materialized because pollsters were wrong about how many pro-choice women and young people would turn out for a midterm election.

Accurately predicting turnout is getting harder and harder as we are undergoing a Trump-related political realignment in this country which nobody has yet completely figured out. And then, even when you do have what you think is a good guess as to who is going to vote, certain demographic groups are almost impossible to reach. Whereas 25 years ago pollsters reached about 40% of the people they called, they now reach 1–2%, so getting an accurage demographic representation has become extremely difficult. Although, ordering Taco Bell delivery has become much easier. So you have to take the good with the bad.

A word about the aggregated polls. These are the polls of polls where groups like 538.com or RealClearPolitics take all the polls and smush them into one big number. This can be useful in the sense that it smoothes out some of the differences between the divergent guesses individual pollsters make. However, there is one thing that makes these polls utterly worthless and something to be avoided.

Republicans figured out something important (they had a special right-handed committee working on it). If your candidates are doing better in the polls, particularly the aggregated polls, it gives them the appearance of momentum, which makes it easier to raise money, get volunteers and drive turnout. So the Republicans established a whole bumch of Republican pollsters who take polls, but make the assumptions we’ve been discussing in a way that will ensure their candidates have better numbers.

So, for example, polling firms like Rasmussen, Trafalgar, Epoch and others might choose to assume that rural white men without college degrees, a group that votes heavily Republican, will make up 12% rather than 10% of the electorate. They then publish results which are very good news for Republican candidates up and down the ballot. These results are then often added to the aggregates along with mainstream, neutral polling, skewing the total aggregated results more Republican. You read these aggregated results, and soon your corduroys are drenched with beer and urine.

So, for all these reasons and more, you should avoid the polls and just focus on working to elect the candidate you like. In the presidential race, there have been 3 national polls by respected neutral pollsters in the last week. All three of them have Kamala with a 3 point lead. We don’t know that these polls are right. They may have each made the same or different incorrect assumptions about who is going to vote. But we do know that the race hasn’t fundamentally changed at all since August.

The only real putpose of polls is not to predict what the result of the election will be. It is to see if there are trends over time. If the same poll, taken every week, makes the same assumptions about turnout. And it has Harris leading by 5, one week, then by 3 the next, then by 1, then down 2, that might be a cause for concern. But that isn’t happening. So go knock doors, make phone calls, donate money, and be helpful to your left-handed friends as they try to negotiate the basic tasks of everyday life.