Why the pollsters obtained it fallacious within the 2016 and 2020 elections and what they may have missed this time

After legal challenges and weeks of recounting votes, the dust settled on the 2020 election with a narrow Joe Biden victory by roughly 40,000 votes in a handful of key states.

It was a far cry from the drubbing of Donald Trump that polls had predicted.

Major presidential polling averages had shown that then-candidate Joe Biden was ahead of then-President Donald Trump by a staggering seven to eight points on Election Day. 

Individual polls were more egregiously faulty, with some showing Biden could be expected to cruise to victory by a 9, 10 or even 11-point margin.

Those forecasts turned out to be wildly wrong. 

Had Trump secured roughly just 10,000 more votes in Arizona, 12,000 more in Georgia and 20,000 more in Wisconsin during the 2020 race, he would have won a second term.

Those totals represent a fraction of a percent separating the Democrat from the Republican, while polling Election Day showed major leads for Biden. 

Simply put, the race was ultimately much closer than polls let on, which naturally raises the question: Are the polls to be trusted in the 2024 election?

Here DailyMail.com breaks down what happened in 2016 and 2020, and what to expect from the polls this time. 

Presidential Polling in 2020 

Pollsters will not beat around the bush, they’ll tell you that they got it wrong last election.

The Head Pollster at Rasmussen Reports, Mark Mitchell, told DailyMail.com that though 2020 proved to be a difficult polling year, some mistakes were more obvious than others.  

‘It was a tough election to understand what was going on, but what you saw was pollsters absolutely clowning themselves with Biden plus 12 and Biden plus 13,’ Mitchell said. 

He said polling that year ‘was a little bit different’ and admitted their results were ‘kind of all over the place.’

Rasmussen’s final prediction ahead of the 2020 election was that Biden had a 1 point lead nationally. That prediction would later prove to be 3.5 points off, though other outfits had much worse predictions. 

According to the final 2020 FiveThirtyEight presidential polling average before the election that year on November 3, Biden led in Arizona by 2.5 points, in Georgia by 1.2 points and Wisconsin by a whopping 8.4 points.

FiveThirtyEight also showed Biden up 8.4 points nationally just ahead of the election.

Clearly, that was not reflective of the outcome. 

But they were not the only national polling average that was off that year. 

The RealClearPolitics (RCP) presidential average showed Biden up 7.2 points nationally on Election Day 2020. 

But Biden ultimately won with just a 4.5 point national advantage. 

In key states he won Arizona by 0.4 points, Georgia by 0.3 points and Wisconsin by 0.6 points.

Election week surveys from CNBC/Change Research, Economist/YouGov and NBC News/Wall Street Journal all found Biden had a 10-point margin heading into Election Day. 

One poll from Quinnipiac University from the same time frame found Biden leading by 11 points. 

The head of the polling program there told DailyMail.com they have since changed their methodology. 

‘Regarding adjustments in methodology, analysis of our 2020 elections results found that an unusually high number of respondents refused to provide an answer when asked which candidate they’d support,’ Director of the Quinnipiac University Poll Doug Schwartz said in a statement. 

These polls were wrong by about six to seven points, and the final results were well outside their margins of error. 

Presidential Polling in 2016

In 2016 the polls faired better, though they still failed to predict Trump’s victory. 

Most surveys found that Hillary Clinton held a couple point lead heading into Election Day. 

This time FiveThirtyEight’s national presidential polling average showed the Democrat with a 3.9 point advantage. 

The RealClearPolitics national presidential polling average showed Clinton up 3.2 points. 

She ended up winning by 2.1 percent nationally, which was in the ballpark of many polls at the time. 

However, what was not predicted was a tide of red voters in swing states that year. 

‘In 2016 the polling was wrong,’ Michael Bailey, a professor with Georgetown University’s Department of Government and McCourt School of Public Policy. 

‘But you know, in the aggregate from the national level, the polls said roughly, Clinton would win by four or something,’ the professor told DailyMail.com. ‘That’s pretty good, but of course, the real problem was in the Midwest, where there wasn’t a lot of polling.’

One issue that year was that pollsters did not get enough white voters without a college degree to answer their surveys, Bailey said. 

‘It’s really hard to reach certain groups,’ he admitted, adding ‘non-college white people’ were ‘really, really hard to get on the phone.’ 

The RCP average for Wisconsin showed Clinton with a 6.5 point lead. In reality, Trump won the state with a 0.7 point margin, meaning the average was a full seven points off. 

Michigan was thought to be solidly in Clinton’s camp too ahead of the 2016 election. She held a 3.6 point lead over Trump on Election Day, per RCP’s average. 

Trump won there too, and the average was nearly 4 points off.

In Pennsylvania that year, the RCP average showed Clinton up 2.1, and Trump won there by 0.7, another nearly 3 point miss by the polls. 

These critical misfires in battleground polling set the stage for Trump’s 2016 victory that shocked many around the country. 

Presidential Polling in 2024

After the consecutive misfires on presidential polling in 2016 and 2020, the industry undertook some self-reflection to identify what went wrong. 

Post-election reviews found that most surveys those years leaned too far left. 

‘93% of national polls overstated the Democratic candidate’s support among voters, while nearly as many (88%) did so in 2016,’ Pew Research found in a review of the two election cycles’ polling. 

Another analysis of the polling in 2020 by industry experts found that pollsters understated Republican support in not only the presidential election but Senate and gubernatorial races as well. 

The panel found that 2020 polls overstated Biden’s support by 3.9 points in the national vote surveys two weeks before that election. 

Understanding that the last two presidential elections were not forecasted as accurately as desired, pollsters from several firms told DailyMail.com they have implemented changes, though not sweeping reforms. 

Schwartz, the lead pollster at Quinnipiac University, said that after 2020 they adjusted their process. 

He said: ‘We implemented a stronger follow-up for respondents reluctant to say how they plan to vote. This tweak seems to be working well as we had one of our most accurate elections ever in the 2022 midterms.’ 

James Johnson, co-founder of J.L Partners, DailyMail.com’s pollster for the 2024 election, said his group and others have also made changes, though not many. 

‘People have made small variations here and there, but they’re pretty inconsistent and they’re pretty different depending on the polls that you speak to,’ Johnson said.

‘The most inaccurate polls in both of those cycles were online only polls,’ he explained. 

‘Someone signs up online, they complete a survey, they get like a little reward for it, like $2 or whatever. The problem is is that those panels tend to be more higher educated, younger and also more likely to be working from home because they’re completing something online, usually during working hours.’ 

This minimized how many true Trump voters were surveyed, he said. And so to account for this J.L. Partners enhanced their ‘mixed method approach.’ 

This means that they incorporate a variety of polling outreach including calls, SMS text messages, and in-app polling to find normally hard-to-reach voters. 

Mitchell of Rasmussen said that this year his operation is closely considering ‘recalled vote’. In polling the ‘recalled vote’ is what respondents say when they admit who they voted for in the last cycle. 

If a pollster, in this case Rasmussen, has an outsized number of respondents claiming to have voted for Biden in 2020, they will adjust the polls to match what the popular vote was so that there are not too many Biden voters represented. 

This method helps make the polling results more reflective of actual voter sentiment, rather than just party affiliation, Mitchell argued. 

‘A lot of other pollsters are not doing that,’ he said. 

The Rasmussen pollster also warned that political leanings may be clouding results. 

‘If the industry has a moderate right leaning polling error, then Harris wins every single swing state,’ he said. ‘If it has a moderate left polling error, then Trump wins complete sweep.’

‘And I think everything I’ve seen points to the latter.’ 

Though changes to the methodology does not necessarily mean the polls will be more accurate this election.

‘It could be that we’re all missing a new dynamic now in Harris’ favor that would be, you know, perhaps suburban women, perhaps Hispanic women,’ Johnson said.

He warned that adjustments made to one part of a poll could end up skewing the results in the other direction, and that the industry still is prone to errors no matter how many changes they’ve implemented.

‘There could be any number of things that we’re missing this time that the polling industry is missing.

‘Applying a lens of the past rather than something new, that’s just a big unknown with polling.’ 

He said there was  ‘one time we’re gonna find that out, and that’s afterwards.’