How Well Can Polls Predict Who'll Win The Election?

The next few weeks will tell us a lot about the state of Joe Biden's campaign against Donald Trump.

Polling on the 2020 presidential race has been pretty unambiguous: Joe Biden is ahead.

Throughout the year, the former vice president has consistently maintained a lead over President Donald Trump ― one that has appeared more robust than Hillary Clinton’s polling lead in 2016. Since June, Biden’s average share of the polls nationally has hovered close to the 50% mark, with Trump trailing in the low 40s.

Being ahead in the polls ― as gun-shy Democrats, indignant Republicans and hedging-prone pollsters all know ― isn’t synonymous with “going on to win.” Surveys serve as snapshots of the state of a race, not predictions of the final results.

Election polling, while not unerring, does generally correspond to the eventual outcome by this point of the race. Most candidates leading at this stage of the campaign go on to win, especially when they’re facing off against an incumbent president. The only incumbent since 1940 to pull out a win after lagging in early summer polls was Harry Truman, according to CNN.

“Historically, we know a lot of what’s going on by this point in time, but we don’t know everything,” said Christopher Wlezien, a University of Texas at Austin political scientist.

Wlezien, along with co-author Robert Erikson, wrote ”The Timeline of Presidential Elections,” a book that tracks the increasing predictive power of polling over the course of a campaign. In January of an election year, they found, polls are almost meaningless as predictive tools. By mid-August, polls tell “about two-thirds or three-quarters of the story of Election Day,” depending on the timing of the conventions.

The next few weeks will be a crucial test for the durability of Biden’s lead. Convention season is a volatile time in the campaign cycle, carrying with it the potential for unusually bouncy poll numbers, either from temporary changes in sentiment or deeper, more fundamental shifts in opinion. Polls rapidly start to become more predictive after those convention effects settle, Wlezien said.

Democrats are understandably wary about getting their hopes up about November after what happened in 2016, when polls showed that Clinton led over Trump. In that campaign cycle, surveys were reasonably accurate nationally, but struggled badly in battleground states. Many pollsters have since implemented a number of reforms, including changes to how they weight polls for educational composition.

What pollsters can’t fix, of course, is the potential for public opinion to change over the course of a campaign, or for late-deciding voters to break hard in the direction of one candidate, as happened four years ago. That leaves a different question: How likely is it that we’ll see a shift in the remaining time before Election Day?

Fundamentally, there’s still a lot of uncertainty about the precise level of volatility we should expect over the remainder of this year’s campaign. There are good reasons to believe the polling now is more predictive than it has been in the past, but the pandemic also throws in an unexpected level of uncertainty. Read on for a look at some of the factors at play.

Why This Year Could Be Less Volatile

Voters are deciding earlier. Traditionally, many Americans haven’t really started tuning into campaigns until autumn rolls around. But, as the Voter Study Group’s Robert Griffin explained last year, polling in recent election cycles suggests that voters are making up their minds earlier, and growing more likely to stick to the decisions they made. Those trends, should they hold in this election, would mean that earlier surveys hold more predictive power than they did in past election cycles.

The electorate may be more stable than it was even four years ago. Comparing recent national polls with those from the summer of 2016, political analyst Kyle Kondik finds, there are now fewer voters who say they’re undecided or that they intend to support a third-party candidate. That suggests there’s somewhat less room for the polls to shift in the same way.

The candidates are known quantities. Both Trump and Biden have experience on the national stage and near-universal name recognition. And a majority of the electorate has strongly held opinions about each man. In one recent poll, 60% of voters said they felt strongly one way or another about Biden, and 77% said the same of Trump. By comparison, at a similar point in 1988 ― an oft-cited example of a candidate’s summertime lead evaporating ― only a third had strong feelings about Michael Dukakis, and just 28% reported feeling strongly about George H.W. Bush.

Trump was, of course, well-known in 2016, too. But now he’s an incumbent president, in a race largely seen by voters as a referendum on his performance. How likely is it that voters’ assessments of him will substantially change?

Throughout his presidency, Trump’s ratings have remained both unusually stable and, continuing a trend set by his predecessors, polarized to an historic degree. Even in the midst of an economic collapse and a pandemic, and despite his poor marks for handling the latter, Trump’s approval rating has remained within a roughly 7-point band. That suggests there’s limited potential for a complete overhaul in voters’ opinions, although whether Trump’s at the bottom or top of that range probably makes some difference.

Early voting is increasingly common. A record share of American voters ― roughly three-quarters ― will be eligible to vote by mail this year. With the pandemic looking unlikely to disappear anytime soon, this year could see an unusually high proportion of voters casting ballots long before Election Day. Voting is set to start as early as September.

A prolonged period of voting also may dampen the impact of a potential “October surprise” in the race. Political campaigns’ effects on voters tend to be relatively short-lived. If voters are casting their ballot over a period of weeks or months, any single event is primed to influence only the smaller group of people who happen to be voting immediately afterward.

Why This Year Could Be More Volatile

Close elections are more common than they used to be. As FiveThirtyEight’s Geoffrey Skelley notes, the 2016 election marked the eighth consecutive presidential election to end with the popular vote winner ahead by only a single-digit margin, the longest stretch since the Civil War. This is the flip side to the idea of entrenched, highly partisan voter preferences ― big, game-changing swings in the campaign may now be less common, but even smaller shifts could be more likely to affect who actually wins.

We’re in the middle of a pandemic. Coronavirus has dominated this election cycle, eclipsing other campaign stories and personally affecting pretty much everyone in the country. Other issues at the top of voters’ minds, like the economy and health care, are inextricably tied into the pandemic. It also remains a volatile situation. It’s clear that the coronavirus isn’t going away, but we don’t know precisely what the situation will look like in November or whether public opinion on the White House response will remain as negative as it currently is. The precise state of the economy is another unknown, as is the degree to which the abrupt economic shifts this year will impact voters’ thinking.

Campaigning in a pandemic looks different, too, with both jam-packed rallies and door-to-door mobilization efforts scuttled or reimagined. Without modern precedent, it’s difficult to say exactly how those changes will affect voters.

It’s even harder than usual to predict voter turnout. As the election nears, more pollsters will be reporting results among “likely voters.” That requires them to make judgment calls about which people are likely to end up actually voting, relying on factors including voters’ past history of turning out. In the best of circumstances, this determination is somewhat subjective and prone to uncertainty. This time around, with the pandemic upending Election Day as normal, it could be more difficult than ever. That’s without even getting into the possibility that ― as many Democrats fear ― rejected or delayed mail ballots could have an effect on the election. Nearly half of all voters say they expect voting to be difficult this year, according to Pew, up from 15% in fall 2018.


What's Hot