Poll: Sanders, Biden seen as most popular second choices in Dem primary
Pollster: Rasmussen Research has a pro-GOP bias
Rasmussen Reports uses research techniques that make its polls favor Republicans, Ipsos Public Affairs research director Mallory Newall said Monday during an appearance on "What America's Thinking," Hill.TV's new show about public opinion research.
Newall singled out Rasmussen's practice of adjusting results by party identification in arguing that the pollster, which has been touted by President Trump and often has shown him with higher approval ratings than in other polls, favors the GOP.
"I think they tend to be more along the partisan angle, leaning toward the Republicans ... the weight that they put toward partisan self-identification," Newall said, referring to Rasmussen's long-standing practice of adjusting its results by party identification.
In response, a representative from Rasmussen in an e-mail to The Hill noted that it correctly predicted the margin in the popular vote in the 2016 presidential race.
"That's a complaint we often hear from Democratic groups not happy with our polling," Fran Coombs, Rasmussen's managing editor, said of Newall's criticism.
"Obviously we're quite comfortable with our partisan breakdown. Let me remind Ipsos that we got the 2016 presidential race right," Coombs continued. "The vast majority of pollsters did not."
In its final poll of the 2016 race between Trump and Democrat Hillary Clinton, Rasmussen indicated that Trump would lose the national popular vote by two points. In the official government tally, he lost by 2.1 percent.
The final presidential survey that Ipsos conducted in 2016 indicated a 3-point Trump loss.
During the 2012 presidential race, Rasmussen said that 49 percent of its respondents supported GOP candidate Mitt Romney compared to 48 percent who backed then-president Barack Obama. The Democrat ended up winning the popular vote, 51 percent to 47 percent.
Since he was inaugurated, Trump has touted Rasmussen's numbers repeatedly, most recently in an April tweet.
"Thank you to Rasmussen for the honest polling. Just hit 50%, which is higher than Cheatin' Obama at the same time in his Administration," the president wrote.
Asked why he believed Rasmussen's numbers are sometimes more favorable to Republicans compared to surveys done by other firms, Frank Newport, editor in chief of the Gallup Poll, pointed to Rasmussen's use of a "likely voter" screen in its methods, something that most polling operations refrain from doing until closer to election dates.
"Most polls use national adults or registered voters and they use 'likely voters' which screens it down," Newport said on Hill.TV. "That skews Republican because Republicans are more likely to vote than not."
Historically speaking, demographic groups that tend to favor Democrats in recent elections have tended not to vote as often as groups that lean Republican.
"Likely voters are a small subset of registered voters. Many people in this country, particularly in mid-term elections or special elections, don't vote," Newall said.
In an analysis released last year, the Pew Research Center found that less than half of Hispanic- and Asian-Americans who were eligible to vote cast a ballot in the 2016 general election. Black turnout also declined as well to 60 percent of eligible voters. Democrats have fared better than Republicans among all three groups.
Sixty-five percent of eligible white Americans voted that year, according to Pew. People of European ancestry were the only racial group that Trump won, according to exit polls.
Determining who is going to vote this November and later in 2020 is the most difficult question pollsters face, Newall said.
"I think that's the key takeaway," she told host Joe Concha. "How do you get the 'correct' likely voter model and if polls are a snapshot in time, how do you build your model to predict what's going to happen in November on Election Day?"
While adjusting results by party identification is becoming increasingly common, it is still contentious in the polling business since partisan identity is not a fixed characteristic like race or sex. Hill.TV's American Barometer polls conducted by HarrisX sometimes utilize party affiliation to correct sampling errors. A research paper released in January by the Pew Research Center found that Democrats were more likely to respond to online polls than telephone-based ones but declined to endorse or condemn partisan weighting in general.
On Thursday, Rasmussen posted an essay on its site which denounced public opinion surveys in its Political Commentary section which frequently features right-leaning voices.
"Polls are just more media propaganda," the column, written by Denver-based physician Brian C. Joondeph was headlined.
"Remember that more than 90 percent of D.C. journalists vote Democratic, with an even higher number giving to Democrats or liberal-leaning political action committees," Joondeph wrote. "These are the people commissioning the polls and interpreting the results for us. Take it all with a big grain of salt."