ALTHOUGH IT’S STILL too early to call the winner of the US presidency, pollsters are already being criticised by some for inaccurate predictions.
In 2016, polling companies and poll aggregators like Real Clear Politics and FiveThirtyEight were widely criticised for their predictions of a relatively clear win for Hillary Clinton over Donald Trump.
FiveThirtyEight predicted that Hillary Clinton had a 71.4% chance of winning the election and Trump with a 28.6% chance. As we know now, Clinton won the popular vote but Trump won the Electoral College vote, leading him to the White House.
This time around, the polling site had given Biden an 89% chance of winning and Trump a 10% chance.
A journalist with the Atlantic described this year’s results so far as a “disaster for the polling industry and for media outlets and analysts that package and interpret the polls for public consumption”.
Two Politico journalists wrote: “The polling industry is a wreck, and should be blown up.”
Almost all of the polls in 2016 put Clinton ahead of Trump overall and polls overestimated Democratic support in some important states like Wisconsin and Pennsylvania.
A lot of the states in which polls were off the mark turned out to to be the ones crucial to deciding the next president.
We won’t know the overall winner this year until possibly tomorrow, but as most states have been called already we spoke to Irish pollsters to hear their take on the situation with US polls for this year’s election.
‘Difficult to poll’
Speaking to TheJournal.ie, Richard Colwell, the CEO of Red C research which conducts polls and other research in Ireland, said so far the US polls seem to have been half right and half wrong.
“I think there is a really big difference between what the polls say and what’s extrapolated out of those polls to work out Electoral College votes and seats,” he said.
You can’t judge the polls really until all the counts are done.
“I can understand why people are calling for [criticism of polls] in the current age when everything is so instant. The immediate response is that ‘oh the polls got it wrong’,” he added, referring to initial results showing Trump performing very well at the start.
He said that the way people voted this time around is so different to previous elections which likely made it “much more difficult to poll”.
More than 100 million Americans voted ahead of Election Day, according to the US Elections Project watchdog.
Mail-in ballots, ballots deposited in drop boxes or cast at polling stations ahead of time represent more than 72% of the total number of votes cast in the 2016 presidential election.
In the previous election, around 57 million people voted early. Colwell said Biden’s current lead of almost 3% ahead of Trump in the popular vote is “pretty much within the margin of error” of predictions at the moment.
However, he said from what we know so far, “some of the state polls were pretty bad”.
“Some were promising pretty strong wins for Biden which then Trump went on to win quite comfortably,” Colwell said.
Colwell said some states with very tight margins should have been defined as “too close to call” by pollsters, instead of putting a really close number on it.
He said that, based on his own experience with polling, voter turnout could explain part of the reason for some polling outliers.
Whenever you’re running a poll, you have to estimate what you expect turnout to be, based on historical elections and what people are telling you.
The much higher portion of voters casting their ballots ahead of time and not on the day could play a part in this.
Colwell added that it is “getting much harder to poll” compared to around 20 years ago when “people responded much more”.
It’s a lot harder to poll now, but it can be done. The election here and in the UK has shown that.
He said there will always be people arguing that pollsters are “not putting in the effort” to fix errors, but “the last thing they want to do is not get the poll right”.
“Complicated place to poll’
Damian Loscher, the managing director of polling and research company Ipsos MRBI, said the United States is a “very complicated place to poll”.
Ipsos MRBI conducted a poll on behalf of RTÉ, the Irish Times, TG4 and UCD for the general election result earlier this year.
It was based on responses from more than 5,000 people at 250 polling stations across the country after voting on the day of the election.
Loscher said the way voting is conducted in Ireland allows pollsters to adhere “to what would be considered robust, scientific research principles”.
“You can’t have a proper conversation about polling without getting really dull and scientific – it doesn’t really suit media discussion and debate,” he said.
As many people in the US vote ahead of time, particularly this year, this hinders the polling process as pollsters use other methods like online surveys.
Loscher said these surveys don’t represent every member of the population and “don’t adhere to these scientific principles in the same way”.
“We have been doing it the same way for the past 50 years because the science supports that ideology. They have moved to new methodologies – once you do this, there are new challenges to pollsters.
And those challenges mean that the plus or minus 3% doesn’t really apply anymore.”
He added that “we won’t know how the polls did until all the votes are counted”.
He said the concern for pollsters is not necessarily being a few points out, but whether any polling issues are due to a systemic problems or a random error.
He added that the negative reaction to US polls at the last election “unfortunately means that when they’re a little bit wrong this time, it looks like a pattern”.
In terms of the theory of the ‘shy Trump’ voter who doesn’t admit to voting Trump on the day, Loscher said this same concept sometimes applies in Ireland.
“There certainly was an element of the ‘shy Trump’ voter in the same way as we have in Ireland had, on occasion, shy voters.
Whether it was a shy Sinn Féin voter because of associations with the IRA or a shy Fianna Fáil voter because of economic collapse, we have always felt that shy voters were reasonably immaterial… not to the point where it would materially impact a poll getting it right or wrong.
He said that these shy voters would not hugely impact poll results, “not to the point where if a poll missed it by six or seven points, you might blame a shy Trump vote for a small fraction of this”.
“This time around, we should not interpret just how good the polls are… until the full polls are counted.”
What did polls say in 2016?
In 2016, FiveThirtyEight predicted a 71.4% chance of Hillary Clinton winning the election and a 28.6% chance of a Trump win.
In terms of the divide over popular and Electoral votes, Hillary Clinton was projected to win 301 Electoral College votes, well above Trump’s predicted 235 votes.
As a reminder, the first candidate to reach 270 Electoral votes becomes the new president.
In terms of the popular vote, it was predicted to be 48.5% to Clinton and 44.9% to Trump. This was within the 3% margin of error for the actual result – 48.2% of the vote went to Clinton and 46.1% to Trump.
Here’s a look at some of the key swing states, and what was predicted in 2016 and 2020 against the actual result in 2016 and what we know so far from this year’s vote.
Florida
Clinton was projected to win Florida by less than 1% of the vote in 2016, but Trump won this state by 1.2% more than Clinton in the end.
Biden was favoured ahead in the projections for Florida in this year’s election, with a lead of more than 2% over Trump.
However, Trump has taken the state with over 3% more of the vote than Biden.
Pennsylvania
This state was also won by Trump in 2016 with less than 1% more of the vote than Clinton.
The polls were just off on this one, with projections on Election Day saying Clinton would win by almost 4% ahead of Trump.
Projections for this election put Biden with a lead of almost 5% over Trump.
This is a key state and Trump is currently leading by almost 2% ahead of Biden with 88% of votes reported.
Votes for this state could be in by the end of today in the US, which could be the deciding factor for the winner of the presidency if Biden takes the state as is currently expected.
Nevada
This state was won by Hillary Clinton in 2016 with a lead of just over 2% over Trump. The polls got this one right, as it was projected by FiveThirtyEight that Clinton would win the state by less than 2%.
This time around, projections put Biden in the lead with 6% more of the vote.
Biden is currently leading in the state by around 11,400 votes
North Carolina
Trump won North Carolina in 2016 with a winning margin of 3.66%. Projections ahead of time said Clinton would win by less than 1%.
In 2020, projections put Biden with a lead of less than 2% in this state.
It’s still too soon to call, but 94% of votes have been reported so far. Trump is currently in the lead by less than 2%.
Ohio
Trump also took Ohio in 2016 with a significant 8% lead over Hillary Clinton. Polls had predicted he would win the state by 2%.
This year’s projections had Trump in the lead with a very tight margin of just 0.6%.
ABC and other news outlets said at 5am on Wednesday that Trump would take Ohio.
Iowa
Trump took the state of Iowa in 2016 with a lead of almost 10% ahead of Clinton. Polls underestimated this one, with projections showing Trump winning by just under 3%.
In Iowa, 2020 projections show Trump winning the state by 1.5%. The Associated Press has called it and said he will win with a lead of more than 8%.
Wisconsin
Trump took Wisconsin in 2016 with just over 1% of the vote ahead of Clinton.
This narrow win was not forecast in the predictions, with Clinton expected to take the state by more than 5% ahead of Trump.
This year, Biden was projected to take the state with a lead of more than 8%.
The Associated Press has already called this state as a win for Biden, but less than 1% ahead of Trump.
Georgia
Trump won this state by more than 5% in 2016 with polls ahead of time predicting a 4% lead for the Republican.
This year’s projections show Biden with a lead of just less than 1% ahead. The race is very tight in Georgia at the moment as the count continues.
It’s still too close to call with Trump on 49.5% of the vote and Biden on 49.2%.
Arizona
Trump took the state of Arizona in 2016 with almost 4% ahead of Clinton. The projections on Election Day that year said Trump would win by just over 2%.
For the 2020 election, Biden was projected to take the state by over 2.5% more than Trump.
Fox News was the first to project that Joe Biden had won the state of Arizona at approximately 4.30am this Wednesday.
False tweets emerged online saying Fox News retracted this prediction around an hour later.
This was untrue and Fox brought on a member of its decision desk to further solidify this prediction.
Fox has continued to stand by its call of Arizona.
The Associated Press also later called this state as a win for Biden. However, the state still hasn’t been called by the New York Times, CNN or CBC as votes are still being counted.
Trump has disputed the call of Biden winning Arizona.
The Associated Press said it made the decision “after an analysis of ballots cast statewide concluded there were not enough outstanding to allow Trump to catch up”.
With 80% of the expected vote counted, Biden was ahead by 5 percentage points, with a roughly 130,000-vote lead over Trump with about 2.6 million ballots counted. The remaining ballots left to be counted, including mail-in votes in Maricopa County, where Biden performed strongly, were not enough for Trump to catch up to the former vice president.
To embed this post, copy the code below on your site
have your say