Analysis:
Analysis: 2020 is probably a lot more competitive than you think

Biden should be wary of polls showing him dominating the 2020 presidential race.

David Rosenberg ,

Donald Trump
Donald Trump
Reuters

If there is one consistent narrative through both the 2016 and 2020 presidential elections, it’s that Trump appears likely to lose.

Donald Trump trailed Hillary Clinton in both national polling and in battleground states for nearly all of the 2016 election cycle – at times trailing by wide margins – and he is now trailing Joe Biden in most polls, often times by double-digit margins.

The polling numbers were so bad in late June, it prompted speculation the president could drop out of the race.

But amidst the near-uniformity of polls suggesting 2020 will be a blowout year for the Democrats, a handful of pollsters are providing an alternative picture of the race, suggesting an election that is highly competitive – and very much winnable for Mr. Trump.

2016 Polling Blues

If the seeming inevitability of Trump’s defeat to Clinton four years ago one was the theme of the 2016 race, the failure of polling to predict what actually happened was the theme of the election’s aftermath.

Data-crunching forecasters all projected a Clinton victory, ranging from 71% likelihood projected by FiveThirtyEight – which was ridiculed for giving Trump too high of odds for winning – to the 98% chance of a Clinton victory predicted by the Huffington Post and the 99% chance reached by the Princeton Election Consortium.

While the national polling wasn’t far off the mark – with the RealClearPolitics aggregate of polls on the eve of the election projecting Clinton winning by three points (she actually won the popular vote by two points) much of the state polling was off, sometimes by huge margins.

Perhaps most notoriously, pollsters confidently predicted Clinton easily winning Wisconsin, with an average margin of 6.5 points accord to the RCP aggregate – yet Clinton lost the state by nearly a point.

The RCP average showed Trump narrowly winning Iowa by three points (he won it by 9.5), and Ohio by 3.5 (he won it by 8.1).

Even in states where Clinton did manage to win, her margins were far slimmer than polling suggested.

In Minnesota, the final polls averaged +9.5 points for Clinton – though she actually only won the state by 1.5 points.

In its autopsy of the 2016 polls, Pew Research Center, echoing the findings of other similar studies of 2016 polling, found that “across the board, polls underestimated Trump’s level of support.”

The most “likely culprit,” Pew found, “is what pollsters refer to as nonresponse bias,” or, in laymen’s terms, the refusal of certain types of voters to share their thoughts with pollsters – what in Britain was dubbed the “Shy Tory” effect.

The “Winner” of 2016

A few pollsters, however, stood apart from the crowd in 2016, drawing ridicule for their projections before election day – and praise after their predictions came true.

Rasmussen Reports, an agency which had been criticized for an alleged Republican bias in its polls, was spot on in the national polling, predicting Clinton beating Trump in the popular vote by two points (she won by 2.1).

In state polling, a newer agency, the Trafalgar Group, drew praise for being the single pollster to predict Trump’s wins in key states like Pennsylvania and Michigan, and was one of only two pollsters to predict Trumps wins in Florida and North Carolina.

Two years later, Trafalgar again predicted a number of key races – this time senate and gubernatorial races in the midterms – which most pollsters missed, including Florida’s Ron DeSantis win over Democrat Andrew Gillum.

Trafalgar’s performance in the 2016 and 2018 election cycles won it high praise from RealClearPolitics cofounder and President Tom Bevan, who declared it “one of the most accurate polling operations in America.”

Harry Enten, a former senior analyst for FiveThirtyEight, dubbed Trafalgar’s Robert Cahaly “one of the most accurate pollsters in 2016”.

Nevertheless, Cahaly, the Trafalgar Group’s founder and chief pollster, says the agency still has trouble finding acceptance in the media, due in part to its rejection of some traditional polling methods and the proliferation of ‘fake polls’, which leads some outlets to ignore newer agencies altogether.

But Cahaly also says the dismissiveness among some media outlets and pollsters is rooted in a refusal to accept that serious mistakes were made in the polling in 2016, with the belief that the misses that year were an aberration, a one-off fluke.

“They don’t want to admit they did anything wrong,” Cahaly told Arutz Sheva.

2020 Polling

Most pollsters show Biden with a comfortable lead over Trump in this year’s presidential election, not only in the popular vote, where he leads by an average of 8.7 points according the RCP average, but even in the electoral college, where the RCP average of polls now projects that if the election were held today, Biden would win a decisive victory, with 352 electors to Trump’s 186, flipping Pennsylvania, Ohio, Michigan, Wisconsin, North Carolina, Florida, Maine’s Second Congressional District, and Arizona.

Some polls even show Biden tied with Trump or leading in Texas, a solidly Republican state which is essential for Trump’s reelection.

Recent polling by Trafalgar, Rasmussen, and SPRY Strategies, however, suggests that the race may be far more competitive than most polls show.

Florida, Michigan, and Wisconsin are all effectively tied, according to Trafalgar’s polls, while Biden has secured a five-point lead in Pennsylvania.

A new national poll by Rasmussen Wednesday found Biden leading Trump by just two points, 47% to 45%, and a massive survey of 42,000 likely voters by SPRY Strategies which was released this week shows Trump leading in Pennsylvania, Michigan, and North Carolina, and trailing Biden by just one point in Wisconsin.

Minority View Matters

It's easy to dismiss the data of three pollsters as outliers, when so many other agencies consistently reach conclusions that are radically different.

And perhaps the majority of polls are right this time, as opposed to 2016, when they performed so poorly, at least in battleground states.

But it is important to bear in mind that the sheer number polls which back up one narrative don't necessarily mean that narrative of the race is more likely to be correct.

Pollsters often engage in 'herding,' as FiveThirtyEight's Nate Silver has noted, tweaking their turnout models and other factors to get the numbers closer to what other polls have.

Even aside from conscious efforts to reach similar results, pollsters tend to cluster around two different sets of results - one showing Trump polling extremely poorly, the other suggesting the race is extremely competitive - because they follow different approaches.

While the details of a firm's methodology vary from pollster to pollster, few actively work to draw out shy voters, including people who declare themselves to be undecided. While Trafalgar has innovated special techniques for drawing out the true voting intentions of such voters, with some other pollsters (like Rasmussen) now adopting some of their methods, most pollsters eschew such methods.

Instead, other pollsters compensate for the lack of registered Republican respondents by weighting their answers, effectively multiplying each response they are able to elicit. That, critics like Trafalgar's Cahaly, say, can lead to warped results, if certain types of Republicans are more likely to respond, or if respondents don't self-declare as Republicans and hide their voting intentions.

Regardless of which polling method is more likely to predict this year's race, it is important to note that the two approaches tend to cluster around similar results - with some variation of course - or at the very least point to similar narratives regarding the race.

Thus if one approach is followed by the majority of pollsters and points to one narrative, it is not necessarily more accurate than the second approach, simply by the number of pollsters who employ it.

This is part one in a two-part series. Part Two can be found here.




top