Video: #AskOBI: How accurate are exit polls in 2020?


November 12, 2020

Josh Clark, our institute's Political Participation Analyst, explains the problems with the narratives coming out of the 2020 elections about how certain demographic groups voted.


Josh ClarkI’ve been seeing a lot of hot takes already about the election results, about which groups did or didn’t vote for Donald Trump. There’s a lot of jockeying to set the narratives coming out of the election about who is responsible for Joe Biden’s victory and for Donald Trump’s defeat.

Narratives matter. They can set the tone for who has influence or potentially get appointments in a new Biden administration. But also beyond that, election narratives matter because the narratives that come out of any particular election influence who are the constituencies that candidates care most about in the next election, who gets courted, and who gets left behind.

So I wanted to speak with you all about when you see things like “Latino men did this,” “White women didn’t do that,” why you should take all of that with kind of a grain of salt or maybe even not pay attention at all.

Most of these figures are coming from what’s called exit polls. Exit polls are basically a big survey project that involves interviewing voters as they come out of their voting locations. There are a couple of problems with using exit polls to say how sub-populations — say upper-income Latinx, Black women, white men without a college degree — voted.

First, the exit polls never talk to enough of any of these sub-groups to have a truly representative or reliable sample. And the second thing is that exit polls don’t take into account how many people actually ended up voting. They don’t know this information yet. None of us do because the vote’s still being counted. So the calculations that are used to estimate what percentage of these groups voted for Trump or for Biden or anyone else are all estimates.

So those are the problems in a regular year. Those are the problems when most people go to the polls and vote. But this year as we all know things were a little bit different. You saw 65% or so of the voters mailing in their ballots or voting early. Interviewing people at polling places is a really limited sample of all of the people who voted. And we also know that those voters who voted in person on Election Day skewed towards Republican voters. So the exit polls are trying to supplement those in-person interviews at voting locations with phone calls to voters who maybe voted by mail or voted early in person. But those polls are really susceptible to all of the same errors that we saw in pre-election polling where there were big misses this year in Wisconsin, Ohio, Iowa

So the point really is that, you know, exit polls are never particularly good at telling us how population subgroups voted. This year they’re probably really really bad at it.

So why are so many people looking at them? Those are really the only figures we have to work with to answer these kinds of questions that everyone is really interested in. Everyone wants to know with what groups did Biden do better than Clinton did in 2016, were there places where Trump picked up new voters, et cetera. In the absence of other, better data that we’ll really need to be able to answer those questions, the exit polls are the only thing that there is.

Knowing that, that makes it all the more important that we be very careful about which narratives we’re lifting up and we’re especially cautious — and I would say resistant — to pushing any kind of narratives that are based on junk figures like the ones that are coming out of the exit polls at this point.