USC Dornsife Daybreak Poll predicts voting trends

The poll surveys 6,000 eligible voters across the United States and contacts participants every two weeks to observe changes in opinion.

Recent data from USC Dornsife’s Center for Economic and Social Research’s 2020 Daybreak Poll shows stability in voter preferences over the presidential race, with Democratic candidate Joe Biden maintaining a slight lead over President Donald Trump.

The Daybreak Poll, which was first introduced as part of the ongoing Understanding America Study in 2014, is conducted by the Center for Economic and Social Research in partnership with USC’s Center for Political Future and the Los Angeles Times. The poll aims to “track change in voting preference in an election panel over time” and “investigate associations of candidate preference and intent to vote with voter characteristics and attributes.”

This year the poll has so far surveyed approximately 6,000 eligible voters across the country, and the results are updated daily.

“One of the most interesting things is that people have made up their minds,” said Survey Director of the Understanding America Study Jill Darling. “Generally speaking, there are events that happen, things go on in the news and in our old life. Before 2016, they would have swung this race. In our new life — in the new United States — they don’t.”

Bob Shrum, director of the USC Dornsife Center for Political Future, emphasized the team’s efforts to represent the diversity of the American voting pool.

“It is designed to be demographically, geographically, economically and educationally representative of the country,” Shrum said. “People are recruited by zip code…if they don’t have internet service, they’re given internet service.”

Contrary to traditional polls that only gather opinions once, the Daybreak poll contacts participants every two weeks as part of a panel survey. The Center then charts their changes in attitude regarding specific candidates and issues to predict voting trends.

Darling said that some events that people thought “might change minds and hearts” didn’t actually affect voter opinions. However, she said the first presidential debate on Tuesday, Sept. 29 might be a chance for the candidates to reach a broader audience.

“I think a lot of people are in media bubbles,” Darling said. “This is a way to kind of reach outside those media bubbles and potentially make some changes. Whether or not that occurs will be very interesting to watch.”

In 2016, the Daybreak poll incorrectly predicted that Trump would win the popular vote because it oversampled the rural demographic. Darling said pollsters often try to predict and correct “understatements of certain groups and overstatements of others.” Because rural voters are often underrepresented, the Center applied weight to their demographic. However, they overcompensated, which led to a skewed prediction.

“When all the other polls had Hillary Clinton up, everyone thought we were the devil incarnate because we were showing Trump ahead,” she said. “Then after the election, everyone said that we were the only poll that got it right, which was very disturbing to me because we didn’t get it right. We’re polling the popular vote, not the Electoral College vote.”

Darling and her team didn’t realize they oversampled rural voters until after the election. Once they discovered their mistake, they made a correction and were satisfied with their methodology.

Their adjusted results showed a one-point Clinton victory, very similar to her actual 2.1% lead on Trump. They then used their model on the 2018 midterm elections, a result Darling said went “very well.” The poll will run the same model in this year’s election.

Although the poll received a B/C rating from — a website that focuses on opinion poll analysis and politics — Darling said the poll is an “experimental academic poll” and that she’s not concerned about the ranking.

“I understand that we have a history of not having the correct outcome,” Darling said. “We’re not a professional polling organization... I think that the ratings that he gives are based on how close a certain number of polls were to the actual outcome in certain categories, and if we get it right, our grade will climb.”