Should We Trust the Polls?
The Runaround, Vol. 0.9.1
First Pitch
A handful of my students have opted to write for Di5W this semester, and I couldn’t be more proud. Check out the Battleground State series from Lone Star Politics” and Danny Machacek’s series on what drives voter turnout.
What To Know This Week
Kamala Harris and Tim Walz are upping their interview game. But they’re never going to catch JD Vance, the interview king.
In class, we talk about the rules, strategies, and realities of the campaign. Campaign news from the last two weeks has been dominated by reality—things that campaigns can’t do much about. That includes hurricanes, book drops, a new indictment by Jack Smith, war in the Middle East, and the Supreme Court’s reinstatement of Georgia’s six-week abortion ban.
The Ted Cruz versus Colin Allred race in is tightening.
Can You Trust Those Polls? Yes.*
In class this week, we talked about the mechanics of polls. Should you trust them?Here’s what I told my students this week:
Many Americans are still traumatized by the NYT election needle from 2016. That (serious) miss, though, doesn’t mean we should throw polling out entirely. To understand whether we should trust polls, here’s some of what I shared with my students.
Polling firms depend on accuracy to keep their business afloat. So they’re trying! But some of them are better than others at making sure that they have a random sample that represents the American electorate.
As we get closer to the election, you’ll basically only see polls of “likely voters,” which means that the polls leave out people pollsters don’t believe will vote, based on their prior voting record and/or demographics.
You should have low trust in individual polls. Look at the polling averages, instead. Take one most recent national poll that has Kamala Harris up 47% to 40%.
That poll has a four-point margin of error, meaning that Kamala Harris might be up 51% to Trump’s 44% or down 43% to Trump’s 44%.
And in any good poll, the most you can say is that we’re 95% sure that Kamala Harris’s numbers are between 43% and 51%. That means there’s a one-out-of-twenty chance it’s higher or lower!
And all that is before we get to the 13% of respondents to that poll that didn’t pick either of them.
The same applied to battleground state polls and the crosstabs, which is probably where we should be looking anyway.
I’ve given you lots of FiveThirtyEight content here—if you want even more, check out their helpful new podcast:
As for 2016, the issue there wasn’t “shy Trump voters,” but a failure of pollsters to “weight” their responses by education. Non-college-educated voters broke decisively for Trump in 2016.
Interestingly, some pollsters have found that if they reweight their 2016 polls for education, they do pretty well.
If the polls are off the mark in 2024, it will likely be because another group—maybe union members or minority voters—break in a similarly surprising way.
A parting shot: trust* the polls, but not the forecasts.
Lagniappe
Did JD Vance’s debate win matter? I don’t have data for you here, but I think so.
Most undecided voters are not deciding between Trump and Harris; they’re deciding whether to vote or not. JD Vance’s seeming reasonableness in the VP debate might mean that right-leaning voters turned off by Trump might be willing to pull the lever for a Republican ticket that includes a VP that looked like an old-school Republican for one night.
Check out the latest research from Pew, Who Americans Follow On TikTok.
Are you focused by the numbering system in these newsletters? Me too. I didn’t want a “Volume 1” until I knew what this was going to become. I still don’t.