Polls are a way of summarizing. When we say that Hillary Clinton is winning (as she is in the most recent Post-ABC News poll, by 4 points among likely voters), that certainly doesn’t mean that no one is supporting Donald Trump. When we say that Donald Trump is winning in Georgia (as he is, although by a narrower margin than past Republicans), it doesn’t mean that everyone in the state plans to cast a Trump ballot in three weeks.
You know that, of course. And yet when the descriptor is more personal — how white men plan to vote, for example, or women with college degrees — polls can feel more personal, and more wrong. Maybe you’re a white man with no college degree who backs Clinton. You’re in the minority among men who share your demographics, but polling doesn’t say that no one with your demographic characteristics exist. Or maybe you’re a young black woman who plans to vote for Trump. Same deal: Polls suggest that you’re not the norm, but that’s it. It works the other way, too. Just because you and all of your friends share similar demographics and presidential preference doesn’t mean that no one like you doesn’t.
Let’s make this more obvious.
We created a tool that overlays polling results from our most recent survey over your own personal demographics. Because sample sizes get smaller as you dive deeper into demographic groups (in other words, we surveyed more men in total than white men who have graduate degrees), we are limited in how predictive we can get from the poll numbers alone. So we created a simple calculation that uses what data we do have to estimate where you might land on the Clinton-to-Trump scale. Bigger circles on the spectrum below indicate larger sample sizes within the poll, and therefore lower margins of error.
The point is precisely that we expect to get a lot of these predictions wrong. And, at the same time, that we can lay bare what our poll tells us about various demographic groups as Americans start to head to polling places.
Of course, you don’t have to plug in your own demographics. See what moves the needle, and how much. There’s a lot that goes into presidential decision-making, and demographic toplines only tell part of the story.
It’s also a reminder that polls deal with margins of error. Polls are upfront about how much they expect to be wrong. There are always people that defy the rest of the demographic group, and there are always estimates that under-represent support in one way or another due to the polls’ methodology.
Broadly, though, polls get it right. Hillary Clinton is winning nationally; Trump is winning in Georgia. That can change by Election Day, but for the most part, we know who will vote, and how.
Even if we guess wrong about you.