Polling in a pandemic: the ideal methodology

Before the coronavirus arrived, it was already likely that more people than ever before would vote by mail, early or absentee this fall. The shift away from in-person voting has happened for decades. Less than 60% of the electorate voted in person in 2016. That decades-long rise in advance voting led us to work with NORC at the University of Chicago to create AP VoteCast. Unable to meet a representative sample of voters standing outside polling places with pencils and clipboards, we built a methodology to meet voters where they are. Regardless of how or when Americans vote, we’re able to capture their opinion via our mail, phone and online approach. Then came the pandemic, which has dramatically accelerated the transition away from in-person voting on Election Day itself. In the 2020 presidential primaries, three times as many people in Ohio, Georgia, Nevada and Virginia voted before the day of the election as in 2016. Five times as many did so in Wisconsin, while advance voting was up by a factor of 12 in Pennsylvania. Come this fall, five states will conduct their election almost entirely by mail. Others will, for the first time, mail every registered voter an absentee ballot. It’s possible —  if not likely — that for the first time ever, most voters will make their choice and cast their ballots before polls open on Election Day.

We didn’t envision a global pandemic when developing AP VoteCast, but that work means one thing AP doesn’t have to do this year is scramble to invent a new way to survey the electorate. We came into 2020 ready for this moment. Here's another thing that makes us excited about using AP VoteCast this fall: our methodology doesn’t require we guess what percentage of the electorate will vote in advance of Election Day. Instead, we ask registered voters to tell us about their plans. We know from our past results that — even in states where changes due to COVID-19 make pre-election estimates difficult or impossible — AP VoteCast will provide an accurate accounting of who will vote when. AP VoteCast’s estimates of advance vote were spot on in the 2018 midterms, with an average error of less than 2 percentage points. That’s going to be an important data point for our decision desk after polls close on Election Day and we start the work of declaring winners. We’ll have that data it for every state, too, which is yet another way AP VoteCast is perfect for this moment. In 2016, not many people went into Election Day thinking the outcome would be decided based on who won Wisconsin, Michigan and Pennsylvania. This year, those states might again be decisive. Or, maybe Arizona and Georgia will matter more. Could Texas be in play? Because AP VoteCast delivers election research about voters in all 50 states, we won’t have to guess which states will make the difference this year. No matter where the “Road to 270” takes us on Election Day (and the days after), AP’s election survey will make sure our newsroom and our customers have an accurate and exceptionally detailed picture of who voted and why.

Learn more about AP VoteCast.