Posted in Behind the News

Understanding the election: Precise polling with AP VoteCast

, by Patrick Maks

AP VoteCast, the wide-ranging survey of the American electorate, wasn’t built with a pandemic in mind, but in many ways uses the ideal methodology to conduct accurate research about the electorate at a time when more Americans than ever before are expected to cast ballots before Election Day.

AP Director of Public Opinion Research Emily Swanson explains how AP VoteCast works and the importance of transparency in polling:

AP conducted a decade of research on the best way to accurately survey the electorate prior to the launch of AP VoteCast. What did we learn and how did that shape AP VoteCast’s creation?

The way that America votes has changed rapidly in recent years. It used to be that the vast majority of Americans who voted went to the polls on Election Day and cast their ballots then and there. In 2016 and 2018, more than 40% of voters voted early – either by mail or by early in-person voting – and that was all before the coronavirus pandemic.

AP VoteCast, which was designed with that shift in mind, meets voters where they are, by conducting AP VoteCast by mail, phone and online, using an approach designed for how America votes today and how it will increasingly vote in the future: early, absentee and by mail. It is no longer the case that going to polling places on Election Day with pencils and clipboards in tow leads to an accurate sampling of the electorate.

How and where are voters contacted? How many people are interviewed?

There are a few different ways we reach out to voters. We send out postcards, which contain a secure link to take an online survey. Each postcard also lists a phone number in case a voter would prefer to call us and conduct the survey by phone. If a voter does not respond to the initial postcard, we follow up directly by phone. Finally, we also include surveys conducted using online panels. AP VoteCast is conducted in all 50 states and will include a total of 140,000 interviews in the general election. For example, in 20 states, we will include 4,000 interviews per state; in another 20 states, we’ll include 2,500 interviews; and in the final 10 states, we will include 1,000 interviews.

Vote stickers are seen at a satellite election office at Temple University's Liacouras Center, Sept. 29, 2020, in Philadelphia. (AP Photo/Matt Slocum)

How do you ensure AP VoteCast accurately reflects the electorate?

We take several steps to make sure AP VoteCast reflects the American electorate. First, we start with a sample of registered voters, which is randomly selected from state voter files. This is considered the gold-standard in public opinion research. We then have a weighting and calibration process to make sure that every demographic group is accurately reflected as part of the survey sample, ensuring that the survey is representative of America as a whole. For online samples, we take additional steps to ensure make sure those samples don’t differ from the randomly selected sample of registered voters in unexpected ways. Finally, we calibrate the results to the vote count, which helps make sure the underlying data at the end of the night is accurate.

We want to be forthcoming about our process because we believe transparency in polling is critically important. Our methodology is available online. We will make our research available for peer and academic review after the November election like we did after the 2018 midterms. Results from AP VoteCast’s research in the 2020 Democratic primaries will be available for review this month.

Swanson explains how AP VoteCast works in this video:

AP announced today where and when it will deploy AP VoteCast in November.

Read more about AP VoteCast.