CIRULI: Fake polls — just a Trump put-down or a real problem?

Author: Floyd Ciruli - October 3, 2017 - Updated: October 3, 2017

Floyd Ciruli

At a recent press conference, Sarah Huckabee Sanders brushed back a question from a CNN reporter about a Fox News poll that showed 56 percent of the American people saw President Trump as “tearing the county apart.” She used Trump’s favorite put-downs:

“A lot of those same polls told you Donald Trump would never be president, and he’s sitting in the Oval Office as I stand here, so I don’t have a lot of faith in those polls.”

She then quoted a poll she liked about support for tax reform. Some polls are fake, others useful.

Listening to Sanders or Trump, you would believe all polling in the 2016 election was a disaster and entirely baseless. Clearly, the narrative going into Election Day created an expectation that turned out to be wrong. But the polling itself was mixed, with most state and national polls accurately capturing the final results. It is important to establish what happened in 2016 and correct any mistakes, as polling has become an essential element in protecting democracy in the Trump era.

Election night

Although Election Night started calmly, it quickly sent news anchors scrambling to explain the unexpected results from early states such as Florida and North Carolina. CNN Anchors Wolf Blitzer and John King, along with other newscasters, finally realized the shocking news. At 3:12 a.m., Blitzer announced to the stunned political and media establishment:

“Donald Trump wins the presidency. The business tycoon and TV personality, capping his improbable political journey with an astounding upset victory.” (CNN, Call the Race, Wolf Blitzer, 3:12 am, 11-9-16)

The final polls published Monday and early Election Day by the leading website aggregators, showed Hillary Clinton winning the popular vote within the margin of error. They ranged from 3 percent on RealClearPolitics, to 4 percent at Nate Silver’s 538 site, to Huffington Post’s 6 percentage points. In fact, the average error rate for major polls in 2016 was better than Election Day polls in 2012.

The uniformity and near certainty of the media narrative describing a Clinton win was mostly based on her lead in national polls from the final debate on October 19, 2016, through Election Day. The only doubt cast on her lead took place between the time FBI Director James Comey announced a review of late discovered emails on Friday, October 28, until, on election Sunday, November 6, he declared no change in the previous FBI position of no charges. Late polls, which appeared after Sunday, showed Clinton weathered the Comey storm and was still headed for a win.

Also, a series of forecasts using polling, big data and modern analytics by Nate Silver, the New York Times and many others uniformly predicted Clinton the winner, some claiming a 99 percent certainty (Nate Silver was the most cautious at 72 percent).

The Narrative

What led to the incorrect narrative that Clinton would win? Obviously, the election was historically unique in that she won the popular vote by more than 2 percentage points, yet lost the electoral vote by 77. But contributing to the powerful reasoning of the Clinton win were a confluence of factors that made Election Night such a surprise for both campaigns, the news media and the professional political class.

Factors that contributed to the Clinton narrative:

  • History suggests a lead in the popular vote presumes an electoral win. A 2-plus percentage win in popular vote nearly ensures 270 electoral votes. Of course, polls had Clinton winning by more than 2 percent, and she barely hit that margin. But, U.S. presidential elections blend geography and voters. The larger problem was polarization of the vote, combined with geographic concentration, which made past assumptions and patterns inapplicable.
  • The forecasts missed the weaknesses in some local polling (e.g., few polls, small samples, completed well before Election Day), especially in a couple of states not getting much attention due to past performances; i.e., Wisconsin and Michigan. Final polls in Wisconsin (Clinton 6.5%) and Michigan (Clinton 3.6%) led to inadequate risk assessment. The final Pennsylvania polls narrowed the last week (Clinton 2.1%), but still had Clinton ahead.

In general, polls and analysts underappreciated Trump’s votes. Election results in Iowa and Ohio, swing states that in recent previous elections voted for Barack Obama, went overwhelmingly for Trump, far beyond final poll estimates (Ohio poll: Trump 3.5%, results: 8.1%; Iowa poll: Trump 3.0%, results 9.4%). Clinton was clearly losing Barack Obama voters. Also adding uncertainty to final forecasts was the impact of third-party candidates (attracting the most votes since Ross Perot in 1992) and final deciders (i.e., weak supporters and undecided the last 48 hours, who strongly broke for Trump). Both elements were underappreciated.

  • Finally, unrelated to polling errors, there were missed analyses. Polls showing Trump was perceived as unqualified with a lower favorability than Clinton led to the assumption he could not win. But, that frame missed the desire for change and dislike of Clinton, which led many voters to support Trump, regardless of their reservations. Emotional factors were trumping what appeared as the rational result given the polling evidence.

Reports and analyses fell prey to a confirmation bias. As the Clinton narrative took hold, confirming statements were repeated and contrary evidence ignored or marginalized. Turnout on Election Day in one area that was good for Clinton was scaled-up into a story that her minority strategy nationwide was working, even when compared to 2012 it appeared muted in key areas, such as Philadelphia. A herd mentality took over normal media caution.

Also, a powerful theory lingered from the big data and demographics in the Obama era that suggested the Democrats were the party of ascendance and the country’s shifting demographics benefitted the party. Unfortunately, the analyses missed both the speed of the shift and the dependability of the loyalty of new voter groups to the Democratic label without Obama on the ticket. In addition, the Democratic-identity politics created a counter move of conservative white-identity politics, and Trump used it well.

Recommendation for Media in 2018 and 2020

In future election cycles, the media needs to create a counter-factual space on its decision desks that specifically looks for and argues against the prevailing narrative. Looking back, there was evidence of a possible Trump upset, but it was mostly downplayed by the power of the Clinton narrative. The multitude of errors in the 2016 coverage requires some restructuring of the process. Media professionalism and viewer skepticism demand it.

The 2016 election’s rancor and split result was a challenge for the political class and the entire electoral system, but polling remains an important and even vital tool in the public realm. It is providing timely feedback to Trump and a political system still in disarray.  It may be more essential now than any other time in recent history as a necessary weapon in exposing false narratives — whatever their source.


Floyd Ciruli

Floyd Ciruli

Floyd Ciruli is a veteran pollster and noted political analyst based in Denver. He is the director of the Crossley Center for Public Opinion Research at the University of Denver's Josef Korbel School of International Studies and is an adjunct professor teaching public opinion and foreign policy. Read Ciruli's blog at www.fciruli.blogspot.com