Stop blaming the polls

U.S. survey

By Mariah Quintanilla

Stop blaming the pre-election polls. They told you all they could about Hillary Clinton and Donald Trump.

The pre-election polls based on national surveys accurately predicted what they were designed to predict: Clinton’s popular election win. So why did so few consider the fact that the electoral college had a real chance of turning the numbers upside down? And why did major news media promote blind trust in the polls, making many people believe a Clinton win was in the bag?

“The polls were pretty much right on the spot,” said Timothy Johnson, professor and director of the Survey Research Laboratory at the University of Illinois at Chicago. Things got tricky when pollsters factored in different electoral college voting models to produce a presidency forecast.

“To accurately get the electoral college, you almost need to do 50 surveys at the state level,” said Johnson.

“If you had had the same kind of large samples sizes in the state of Michigan that you had at the national level, you would have probably seen this coming,” he added.

One problem with national pre-election polling is that sample sizes are significantly smaller than for other surveys.

“In pre-election polling, the response rate is often 10 percent,” said Allyson Holbrook, associate professor of social psychology at the University of Illinois at Chicago. Though surveyors try to make the sample sizes as representative of the national U.S. population as possible, it’s hard to capture the infinite social nuances that comprise the American populace.

People trusted the polls to accurately predict their future president. When the polls failed to do so, liberal America felt as though it had been misled, either by the polls, pundits or news media. Did the media sufficiently and accurately warn people of the polling risks involved in a race where popular votes and an electoral college can create a divisive plurality?

Though many electoral college explainer videos became popular after the election, various publications thought it was a good idea to re-educate their audiences about the interworking of the electoral college in preparation for election day.

Al Jazeera broke down the American voting system in an article for a foreign audience looking on before the election. The author tried to  assure people that the U.S. election would not be as unprecedented as Britain’s Brexit vote, when to the shock of many, nearly 52 percent of the country voted to leave the European Union.

“It has become trendy for American political pundits and even some Trump supporters to cite the British “Brexit” vote as an example of a vote where all of the polls were wrong and a right-wing candidate/idea upset the establishment,” wrote Jason Johnson, a professor of political science and politics editor for The Root.

“For a candidate to pull off an upset in the American system the national polls and literally dozens upon dozens of state polls would all have to be wrong and likely all wrong in the same direction. That kind of systematic error is highly unlikely,” he said.

The truth is that most liberal news media were staking their predictions on national polls supplemented by historical trends. In the hyperpolarized political and social unrest of 2016, historical trends lost meaning – and accuracy.

Some news media’s “Las Vegas style odds” and high degrees of certainty may have promoted the notion that there was little to no chance of Donald Trump winning the election, said Timothy Johnson. On October 3, for instance, the Huffington Post gave Clinton a 98 percent chance of winning the presidency.

In the days leading up to this year’s election, everyone from Fox News to CNN ran polls that put Clinton 2 to 5 percentage points above Trump. Nate Silver, an influential poll aggregator and journalist, was criticized for developing one of the most conservative presidential forecasts of the all the major news networks on his website FiveThirtyEight.

On the morning of election day, Silver’s pre-election models gave Clinton a 71.4 percent chance of winning the election, leaving a significant 28.6 percent chance for Trump. In the week following the election, Silver was called to essentially defend his “polling honor” on a variety of talk shows for failing to accurately predict the outcome.

“If the news media had acted like Trump had a 30 percent chance of winning, which is what our forecast said, then I think they would have acted very differently,” he told The Daily Show host Trevor Noah.

The Investor’s Business Daily (IBD) /TechnoMetrica Market Intelligence (TIPP) Presidential Election Tracking Poll and the University of Southern California’s (USC) Dornsife / Los Angeles Times Presidential Election Poll consistently forecasted a Trump tie or lead in the weeks leading up to the election. The IBD/TIPP poll put Trump 2 percentage points ahead of Clinton on election day. And yet, their margin of error meant their estimates could have been off by 3.1 percentage points in either candidate’s favor.

In the month following the election, pollsters and surveyors have been continuously blamed for their inaccurate predictions. “The morning after the election, I saw messages saying that election polling as we know it is dead,” said Holbrook. While polls can tell us who people intend to vote for, she added, it’s hard to predict whether those people will vote. Improvements, however, rely on the American public not losing faith in national surveys.

Accurately forecasting future elections may consist of improving pre-election polls by providing a more representative and nuanced sample of the U.S. population, funding state-level polls that would better estimate the electoral college vote, and convincing more people to participate in surveys.

The day after the election, senior researchers at Pew Research Center attempted to assure the public that polling is not dead, only in a state of essential experimentation.

They wrote, “The role of polling in a democracy goes far beyond simply predicting the horse race. At its best, polling provides an equal voice to everyone and helps to give expression to the public’s needs and wants in ways that elections may be too blunt to do.”

Photo at top: The sample size for national election polls tends to be smaller than for other kinds of surveys. (Pixabay)