Analysis

Why Do Experts Keep Getting Polls So Wrong?

(Photo by JEFF KOWALSKY/AFP via Getty Images)

Michael Ginsberg Congressional Reporter
Font Size:

Political operatives have struggled to analyze polls over the last decade, and the problem may be getting worse.

Polling in the 2020 Senate races of key GOP incumbents like Maine’s Susan Collins, Iowa’s Joni Ernst, and South Carolina’s Lindsey Graham consistently showed Democratic challengers running closely against the Republicans, although all three Republicans won by at least six percentage points. One poll, conducted less than two months before election day, showed Collins losing to Democratic challenger Sara Gideon by 12%, although Collins ultimately won by nearly ten points. (RELATED: Democratic Candidates In 12 Competitive Senate Races Raised Over $275 Million In Three Months)

Pollsters ultimately underestimated voter support for former President Donald Trump and Republicans in the House of Representatives by four percentage points. While the presidential polling error was in line with 2012 numbers, the error in House of Representatives elections was pollsters’ worst since 2006, according to FiveThirtyEight.

In the run-up to the 2020 election, many pollsters stressed the importance of weighting for educational attainment. Weighting is the process by which pollsters make sure that the groups they poll have a similar demographic composition to the overall electorate.

“People with higher levels of education are more likely to participate in a poll, to answer the phone, to fill out a survey. Most national pollsters knew about this and had a protocol in place. But that practice was not being done by most organizations doing state polling” in 2016, Courtney Kennedy, director of Survey Research for the Pew Research Center, told NBC News in July 2020.

Over the last election cycles, education polarization has increased, with highly-educated individuals becoming more likely to vote for Democrats, and individuals with less education becoming more likely to vote for Republicans. As a result, unweighted state polls showed bias towards Democrats.

Other pollsters changed how they contacted potential voters, hoping to reach a broader slice of the electorate.

“We find that younger people, but also men and people in urban areas, would much rather answer a poll by text than be called on the phone,” Tom Jensen, the director of Public Policy Polling, told FiveThirtyEight.

In addition to their over-representation in polling, younger, more-educated, and urban individuals are over-represented among Democratic Party operatives, according to data analyst David Shor. As a result, Shor said, the Democratic Party has struggled with working-class, less educated, and more rural voters, while failing to understand why.

“People who work in campaigns are extremely young and much more liberal than the overall population, and also much more educated,” he told Politico. “I think that this is pushing them to use overly ideological language, to not show enough messaging or policy restraint and, from a symbolic perspective, to use words that regular voters literally don’t understand.”

As a result of these errors, both pollsters and the Democratic Party over-estimate the number and electoral importance of college-educated, urban voters. In contrast, when Shor worked on former President Barack Obama’s successful 2012 campaign, he explained, Democrats relentlessly messaged on popular left-wing economic issues, and lost non-college educated white voters by 18 points. That margin still was enough to put Obama over the top in swing states Pennsylvania, Wisconsin and Michigan. Non-college educated white voters supported Trump by 36 points in 2016, giving him those three key states. President Joe Biden lost that group to Trump by 26 points in 2020, but that was still enough to swing Pennsylvania, Wisconsin and Michigan back into the Democrat column.