Top Gallup editor: Faulty polling led to inaccurate 2016 data models

Frank Newport, editor-in-chief at the Gallup Poll, on Monday said that data analyst websites, not national polls, were the ones incorrect in their 2016 presidential election predictions. 

“The national polling was actually one of the more accurate years where they correctly predicted the popular vote,” Newport said to Hill.TV's Joe Concha, referring to national polls. 

“But one of the things that happened was an overreliance on these models where various individuals and various companies put together these scientific-sounding probabilistic models,” he continued. 

“When you disassemble those models, all they were were looking at state polls,” he said. 

He pointed at the smaller sample sizes in toss-up states making it difficult to get accurate polling.

“In Minnesota for example, the most recent poll before the election was eleven days old ... and those state polls are very tricky anyhow because it's a smaller sample, and small differences in turnout between big cities and rural areas could make a big difference," Newport said. "Those models sounded scientific, but they were relying on inferior polling at the state level in these difficult to poll states that are toss-up states.”

Newport was referring to popular analyst websites such as FiveThirtyEight, the New York Times' UpShot, and the Huffington Post's Presidential Forecast, all of which predicted Democratic nominee Hillary ClintonHillary Diane Rodham ClintonGraham: There's a 'bureaucratic coup' taking place against Trump Fox News poll shows Dems with edge ahead of midterms Poll: Democrats in position to retake the House MORE would defeat President TrumpDonald John TrumpGrassley: Dems 'withheld information' on new Kavanaugh allegation Health advocates decry funding transfer over migrant children Groups plan mass walkout in support of Kavanaugh accuser MORE

The websites faced widespread criticism for their predictions following the election. 

FiveThirtyEight founder Nate Silver last year said that the blame should be on the news media for cherry-picking data that supported their belief Clinton would win and not fully grasping how data models work.

“Journalists just didn’t believe that someone like Trump could become president, running a populist and at times also nationalist, racist and misogynistic campaign in a country that had twice elected Obama and whose demographics supposedly favored Democrats,” Silver said. 

“So they cherry-picked their way through the data to support their belief, ignoring evidence — such as Clinton’s poor standing in the Midwest — that didn’t fit the narrative,” Silver said.

— Julia Manchester