FACT CHECK: Trump Says Rasmussen Was One Of The Most Accurate Pollsters On Election Day
President Donald Trump said in a tweet Tuesday morning that Rasmussen was one of the most accurate polling agencies on Election Day in 2016.
Rasmussen just came out at 51% Approval despite the Fake News Media. They were one of the three most accurate on Election Day. Just about the most inaccurate were CNN and ABC News/Washington Post, and they haven’t changed (get new pollsters). Much of the media is a Scam!
— Donald J. Trump (@realDonaldTrump) April 17, 2018
“Rasmussen just came out at 51% Approval despite the Fake News Media. They were one of the three most accurate on Election Day,” he tweeted.
Rasmussen accurately forecasted that Clinton would win the popular vote by 2 percent, though other pollsters were better at predicting how much of the vote each candidate would receive. The ABC News/Washington Post poll was far more accurate than Trump claims.
Rasmussen polls the American public on Trump’s approval rating nearly every day. It found a 51 percent approval rating on April 16, though the average approval rating for April 16 to April 18 was 48 percent. Leading up to Election Day in 2016, it polled the public on who they planned to vote for in the presidential election.
The 2016 election shocked voters, news outlets and pundits who thought that the polls did not seem to predict a Trump victory.
National polls leading up to the presidential election forecast the national popular vote, not the electoral college outcome and ultimate winner. While Trump won the electoral college, he lost the popular vote. Democratic presidential candidate Hillary Clinton won almost 3 million more votes than Trump and earned 48 percent of the vote to his 46 percent.
In the final polls before the election, many national pollsters correctly and closely forecasted a Clinton popular vote victory. Rasmussen pinned the exact spread: a 2-point Clinton win. The polling company called itself the most accurate pollster measuring the national popular vote in 2016.
No other agency found a 2 point spread in favor of Clinton in a four-way race between Clinton, Trump, Libertarian candidate Gary Johnson and Green Party candidate Jill Stein in its final poll. A McClatchy/Marist poll found a 2-point Clinton win when it asked only about Trump and Clinton, though. Rasmussen did not conduct a two-way poll.
Vote spread is not the only measure of accuracy. Polls can also be evaluated by how close they come to the actual vote breakdown.
“Rasmussen did project the final, major-party spread relatively closely, but it is also fair to point out that they underestimated votes shares for both Trump and Clinton considerably,” Costas Panagopoulos, a political science professor at Northeastern University who rates poll accuracy in election years, told The Daily Caller News Foundation in an email. Several other polling firms came closer to the actual vote shares.
In its polling for a four-way race, Rasmussen underestimated Trump and Clinton popular vote shares by about 3 points. It predicted 45 percent to Clinton, 43 percent to Trump, 4 percent to Johnson and 2 percent to Stein.
Polls like those conducted by Fox News and ABC/Washington Post were closer to the actual vote proportions than what Rasmussen had predicted, though they were a point or two off from the actual vote spread. In two-way polls, Fox forecasted 48 percent to Clinton and 44 percent to Trump, and ABC/Washington Post found Clinton leading 49 percent to Trump’s 46 percent.
It is not clear which other two firms Trump is referencing when he says that Rasmussen was one of the “three most accurate” pollsters, though it is possible he was thinking of the two polls that predicted him winning.
The day before the election, the IBD/TIPP Tracking poll found Trump leading by 2 points in a four-way race, though it showed Clinton up by 1 point in a two-way race. A LA Times/USC Tracking poll forecasted a 5 point Trump lead in a two-way race the day before Election Day. These polls, however, were wrong about the popular vote winner and spread.
Panagopoulos ranked the accuracy of the 2016 presidential election polls and found that the McClatchy/Marist, IBD/TIPP Tracking and ABC/Washington Post were the most accurate. Rasmussen is not on the list because Panagopoulos only ranked two-way polls between Clinton and Trump and Rasmussen only polled for a four-way race.
Trump said that the ABC/Washington Post poll was one of the worst, but it was actually one of the most accurate.
“Rasmussen saying Clinton would win by 2 and ABC saying Clinton would win by 4 are pretty similar findings and both pretty close to what happened,” Kristen Soltis Anderson, a Republican pollster and co-founder of Echelon Insights, told TheDCNF in an email.
CNN did not release a poll in the week leading up to the election, but its poll released on Oct. 24 before the election found Clinton leading by 5 points.
By historical standards, the national polls were generally accurate, and the election result was within the margin of error of many polls. “The real issue with the polls in 2016 wasn’t the national polls,” Soltis Anderson said. “The problem in 2016 was the statewide polls in the ‘Blue Wall’ Midwest states.”
An American Association for Public Opinion Research (AAPOR) analysis of 2016 election polling found that state-based polls had a particularly bad year forecasting state outcomes. For example, Trump narrowly won Wisconsin and its 10 electoral votes, but polls leading up to the election found Clinton leading in the state by 4 to 8 points.
Poll aggregators like FiveThirtyEight and The New York Times’ The Upshot attempted to predict who would win the electoral college by building models that relied on state-based polls that turned out to be inaccurate. A few weeks before Election Day, The Upshot gave Clinton a 93 percent chance of winning the election. “They helped crystalize the erroneous belief that Clinton was a shoo-in for president,” the AAPOR report said.
Differences in poll methodology – such as question wording, whether a poll is a “robo call” or conducted by a live caller, how many people respond to the poll and how responses from different demographics are weighted – can contribute to varied results.
“You have shops like PPP who do robo-polling on the cheap, saying, for instance, Clinton was going to win Michigan by a lot. Or you have local media outlets or universities who are sometimes doing polling on very small budgets,” Soltis Anderson said.
Rasmussen has a mixed history of correctly forecasting elections. Its “final poll of the 2008 general election – showing Obama defeating Arizona Sen. John McCain 52 percent to 46 percent – closely mirrored the election’s outcome,” wrote Politico.
But in 2010, Nate Silver of FiveThirtyEight called Rasmussen “biased and inaccurate,” skewing in favor of Republicans. FiveThirtyEight gave Rasmussen a C+ rating for its slight Republican skew. Panagopoulos found that Rasmussen was one of the five least accurate polls in 2012.
The Associated Press Stylebook announced new guidelines Tuesday related to reporting polls. “Poll results that seek to preview the outcome of an election must never be the lead, headline or single subject of any story,” it says.
Have a fact check suggestion? Send ideas to [email protected]