As the dust settles on the election, it becomes increasingly clear that the margins of victory for former Vice President Joe Biden were razor-thin in a number of battleground states—a far cry from the decisive, near-double digit victories forecasted by Emerson Polling.
Emerson’s final batch of statewide polls, conducted from Oct. 29-31, predicted convincing defeats of the incumbent president in Michigan, Wisconsin, Pennsylvania, and Florida. Other races, like Ohio, Iowa, Texas, and Maine’s second congressional district, were predicted to be closely competitive. Yet, after days of vote counting, the results reported to date by the Associated Press and several other news organizations tell a different story.
Biden eked out a victory in Wisconsin by just 0.6 percentage points, rather than Emerson Polling’s predicted eight points. In Michigan, Biden won by just under three points rather than seven as predicted. And in Pennsylvania, where Emerson Polling foresaw a four-point Biden victory, the now-president-elect holds a lead of just half a point, with 98 percent of the vote reported. While these states were key in determining the presidency, their results fell outside the typical margin of error of three points.
Emerson Polling Director Spencer Kimball said the discrepancy between the polls and the actual results was due in part to record voter turnout in the 2020 elections. Whereas 139 million Americans cast ballots in 2016, this year’s projected turnout of 158 million—66.4 percent of registered voters—would be the largest in any presidential contest since 1900.
“That voting increase was disproportionate around the country,” Kimball said. “So it was hard to account for that unknown quantity.”
Among polling organizations, Emerson Polling is held in high regard, rated the country’s most reliable automated poll service in 2018 by FiveThirtyEight. Nevertheless, its 2016 predictions estimated a landslide Electoral College victory of 323 to 215 for Democratic candidate Hillary Clinton.
In other battleground states favored for Biden, the Democratic candidate lost outright. Emerson Polling had the former vice president leading by one point in Ohio and by two in Maine’s second congressional district. President Donald Trump carried those races by eight and seven points, respectively, far outside of the typical margin of error in state polls.
In Iowa, where the latest Emerson poll predicted a one point Trump win, the president won the state by more than eight points. Texas, labeled a statistical tossup in the latest poll, went to Trump by nearly six points.
One of the largest discrepancies in the group’s polls was its assessment of Florida. Trump carried the state, and its 29 electoral votes, by three percentage points. On Nov. 1, Emerson Polling had Biden leading by six.
Kimball said he felt uneasy about the final Florida polling numbers even before results began pouring in. As it turned out, his projections for Miami-Dade County—which went to the past three Democratic presidential candidates by 15 point margins—failed to anticipate an inflated Republican turnout. On Election Night, he said he considered the state lost by the Biden campaign as soon as the first numbers were reported.
“Those types of mistakes are systemic, where, as methodology, you have to decide what you’re going to do,” Kimball said of his reluctance to publish the Florida poll with its questionable figures. “Because there were so many [other] polls showing [a Biden victory in Florida], for me to throw out that poll would have looked more partisan than polling. That’s what the numbers look like.”
Kimball pointed out the statistical divide between in-person and absentee ballots, rooted in conflicting attitudes towards the COVID-19 pandemic. In-person ballots—which trended Republican due in part to Trump’s unfounded allegations of fraud associated with mail-in voting—made up the initial counts in most states. Only after states began processing mail-in ballots—which conversely trended heavily Democratic—did the results swing towards Biden. Some states, including Pennsylvania, mandate that mail-in votes cannot be counted until Election Day.
“I think [as of Nov. 6] there’s three or four hundred thousand outstanding ballots,” Kimball said of the absentee ballot counting in Pennsylvania, a race called by AP on Nov. 7. “At the time of this recording, they’re only reporting about one, two hundred [thousand ballots]. That number will make a huge difference in Biden winning the state by two points, or by four points.”
Kimball said he was proud of the work that Emerson Polling did in 2020. He cited the organization’s accurate polling in states like New Hampshire, Arizona, and Nevada, and on the national level. Biden won the popular vote by three percentage points, close to Emerson’s predicted five.
“Our national numbers, I think, are going to reflect what really happened,” Kimball said. “It’s hard to change what works. But obviously, our state polling needs to have some attention.”
Emerson Polling would first await the final results of the election before assessing any potential changes to its modeling, Kimball said.
“After 2016, we added to the methodology, [and] in 2018, even though we did exceptionally well, we added to the methodology,” he said. “As communication continues to evolve, so does the way that we collect data. Sometimes in that process, we slip up, but that is in the process of learning. And in communication, it’s always evolving.”
In recent years, Emerson Polling introduced new polling mechanisms, like a new SMS-based initiative meant to supplement web-based and landline polling.
”You can only text in Montana, and our polls were accurate there,” Kimball said.
Furthermore, Kimball said he is confident that polling, as an institution, will survive the aftermath of Nov. 3, 2020—in part due to sheer public demand for raw data amidst the anxiety and uncertainty of election years.
“‘Don’t compare us to the Almighty,’” he said, quoting former Boston Mayor Kevin White. “‘Compare us to the alternative.’”
