Q: What is AP VoteCast?
A: VoteCast is a survey of the American electorate aimed at telling the story of why the winners won.
Based on nearly 139,000 interviews with registered voters in every state, VoteCast delivers data about the makeup of the American electorate nationwide and in all states holding an election for U.S. Senate or governor in 2018. It reveals opinions held by American voters about the candidates, the issues they care about and their views on the future of the country. It’s a survey that powers AP’s race calls on election night and ensures AP and its customers can tell the full story of Election Day.
Q: Isn’t this what AP did with an exit poll?
A: For years, AP was part of the consortium of media companies that conducted election research with an exit poll. An exit poll is exactly what is sounds like: a survey of voters interviewed in person as they “exit” a neighborhood polling place. That approach no longer works for how America votes today. In 1972, when President Richard Nixon was re-elected, 95 percent of voters cast a ballot in person on Election Day. In 2016, 42.5 percent of voters didn’t go to the polls on Election Day, choosing instead to vote early, absentee or by mail. VoteCast moves away from traditional in-person exit polling for this reason.
Q: Who developed the VoteCast methodology?
A: For the past decade, AP has conducted research, funded in part by the Knight Foundation, aimed at uncovering ways to improve the in-person exit poll. This research is the foundation of VoteCast’s modern approach to accurately reflecting the opinions of the American electorate.
To develop VoteCast, AP worked with NORC at the University of Chicago — one of the world’s preeminent objective, nonpartisan research institutions and AP’s partner in the AP-NORC Center for Public Affairs Research. AP and NORC worked with Fox News in 2017 to test a new approach to election research in three statewide elections.
Q: How does the VoteCast survey process work?
A: Starting eight days before and concluding as polls closed across the country on Election Day, VoteCast interviewed approximately 139,000 registered voters in English and Spanish. By comparison, a high-quality pre-election poll is usually based on interviews with between 600 and 1,000 people.
Our questionnaires ask a standard set of demographic questions, such as age, race and ethnicity, marital status and income. We also ask a series of opinion questions, seeking details on what issues voters care most about, their thoughts on the direction of the country and their opinion on a variety of topics that are specific to their state. We also ask for whom they cast their ballot for U.S. Senate and governor, as well as the party of their candidate of choice for U.S. House. In a few states, we also ask about ballot measures and races for statewide office.
We start by mailing a postcard to a random sample of registered voters in 25 states, inviting them to take our survey either online or by phone. We also try to reach those registered voters directly by phone. At the same time, we’re conducting a random-sample survey of registered voters nationwide. Finally, we survey self-identified registered voters in all 50 states using online panels, which allows us to interview a very large number of people in just a few days.
VoteCast then employs an innovative series of calibrations to combine the highly accurate random sample surveys of registered voters with the poll conducted via online panels. In doing so, we get the best of both approaches. VoteCast delivers the accuracy of a survey conducted with a random sample, but also provides the research depth of a survey that interviews tens of thousands.
The final step in this process calibrates VoteCast results with the actual vote count. Because of the large VoteCast sample, we can complete this process in as many as 30 separate areas of a state, further adding to the survey’s ability to accurately reflect the opinions of voters.
More details about the VoteCast methodology can be found here.
Q: What are the primary advantages of VoteCast over other kinds of polling?
A: In addition to its improvements in accuracy, there are two primary benefits to VoteCast’s approach to election research.
Because it is not based on in-person interviews, VoteCast is able to capture the opinions of both people who vote and registered voters who decide not to vote in an election. VoteCast also returns results in every state holding an election for U.S. Senate or governor in a midterm election year, and in every state in a presidential election year. This means VoteCast delivers a broader portrait of the American electorate than in any previous election survey.
And while VoteCast is new, over time it will become a resource to examine how the American electorate has changed. AP is committed to ensuring VoteCast delivers a robust data set in every midterm and presidential election, as well as in presidential primaries, allowing for rich longitudinal comparisons and exceptional journalism.
Q: Why is AP confident in the results it delivers with VoteCast?
A: VoteCast is based on a decade of research and experimentation aimed at improving the in-person exit poll and, later, creating a replacement. AP successfully tested the VoteCast methodology in the 2017 gubernatorial elections in New Jersey and Virginia, and in the Alabama special election for U.S. Senate. At poll close, the surveys accurately forecast the winner and the winner’s final vote share.
That success was repeated in the 2018 midterm elections. In races for Senate and governor, AP VoteCast correctly projected the winner in 92 percent of races at 5 p.m. In the others, AP VoteCast had two as a tossup, with a projected difference between the candidates of less than one percentage point; three races were so close that the final outcome wasn’t decided until more than a week after Election Day; and one incorrect winner was projected.
Our research and development will continue after every election. While we expect the core VoteCast methodology will not change, AP and NORC will learn from this year’s results. We updated the VoteCast methodology following our experiments in 2017, and we expect to do the same before VoteCast returns in the presidential primaries in early 2020.