Comparisons between the surprise outcome in the UK referendum vote earlier this year and the upcoming US presidential election are hard to avoid, but we think the differences outweigh the similarities. Most importantly, polling in the UK was much closer, and much less consistent, than it has been in the US presidential campaign this year.

In our view, the remaining sources of uncertainty do not lean clearly in either direction, but on the margin favor Sec. Clinton more than Mr. Trump. First, although there is a discrepancy between online and live-caller polling, most of the difference is related to Sec. Clinton’s share, contradicting the notion of “shy Trump” voters.

Second, polling suggests that if the large share of undecided or third-party voters shifts, they are more likely to support Sec. Clinton than Mr. Trump. Third, while there are mixed signals regarding turnout, preliminary data regarding early voting in key swing states like Florida and North Carolina suggest that Democrats might make larger gains in turnout than Republicans this year.

We think that the upcoming U.S. election won’t end up as another Brexit-styled surprise for for two reasons.

First, and most importantly, whole both situations represented an opportunity for voters to endorse a change in the status quo, voters in the UK were asked to decide on an idea whereas in the US they are being asked to decide on a person. The distinction is illustrated in US polling by the difference between the small share of Americans who believe the country is moving in the right direction (29%) and majority who approve of the job President Obama is doing (52%).

Second. While the polls conducted on the eve of the referendum vote showed “remain” with a 4.6pp lead, in contrast to the 3.8pp actual vote margin in favor of “leave”, an average of polls published by the Economist magazine the day before the election showed a tied race, and showed “leave” leading for much of the prior month. As much as 10% of the public in many of these surveys was also undecided. By contrast, Sec. Clinton has led the average of presidential polls consistently for more than a year, with the exception of one week in late July following the Republican convention, and for most of the last year her lead has been substantial, averaging 4pp since the last primary elections were held.

Indeed, many observers expected Mr. Trump’s bid for the Republican nomination to fail until the nomination process was more than half complete. But if there is a lesson to be drawn from that experience, it is that one should pay more attention to polls and less attention to one’s own expectations. From shortly after Mr. Trump announced his candidacy in June 2015, he took the lead in the average of national polling on the Republican nomination, and never lost it. Most observers, including ourselves, discounted this, assuming that most of the two-thirds of primary voters supporting other candidates would consolidate around someone else. In this case, focusing mainly on the polls would have been more useful than making assumptions about the future behavior of voters.

More generally, primary elections are very different from general elections, because the former lacks the strong influence of the two-party system. While around 40% of voters are unaffiliated with either of the major political parties, most voters “lean” in one direction or the other; Pew Research has found that less than 10% of voters are truly “independent”. Most demographic groups also support one party disproportionately. This polarization of the public along party lines arguably leads to closer elections—the vote margin for the winner of the last four presidential elections averaged only 3.3pp—but it also provides for a much clearer baseline of support for each candidate than primary elections do.

Over the last few months, polls published by the Los Angeles Times and the University of Southern California (LA Times/USC), Rasmussen, and Investors Business Daily (IBD/TIPP) have shown Mr. Trump to have more national support than most other surveys. The LA Times poll is unusual in several respects, including its repeated polling of a fixed sample of voters, rather than new random samples, its heavy weighting of demographic groups underrepresented in the sample, its use of self-reported 2012 presidential preference in weighting, and its questioning, which asks respondents to assign a probability to their vote for a given candidate. It is unclear what explains the Rasmussen polls divergence, but we note that the poll is estimated to have a 2pp historical bias favoring Republican candidates. The IBD/TIPP poll is more traditional, but recent polls have used a disproportionately Republican-leaning raw sample, which may affect the results even after weighting. We note that while the IBD/TIPP poll was among the most accurate in the final polls leading up to the last three presidential elections, reviewing the accuracy of polls conducted two to four weeks prior to Election Day in 2004, 2008, and 2012 shows an above-average error.

Of course, some outliers showing Sec. Clinton leading by a wide margin may have methodological issues as well; a recent ABC News/Washington Post poll that showed Sec. Clinton leading by 12pp used a sample that included only 27% Republicans, rather than the 32% in the 2012 exit polls, for example. Estimates of turnout among different groups is the most likely source of polling error, in our view, but also what sets good pollsters apart from bad ones. Regardless, these distinctions might matter less now than they once did, as all the most recent waves of all three of these outlier polls are now showing a small 1pp lead for Sec. Clinton.

Mr. Trump does better in automated, more anonymous polling, but mainly because of fluctuations in support for Sec. Clinton across polling methods. Sec. Clinton’s lead is much larger in the average of recent live-caller polls than in the various automated polls (internet, as well as automated phone polls). This pattern was also observed ahead of the UK referendum on the EU, when live-caller polls generally showed greater support for “remain” than internet polls. One theory for this discrepancy, known as “social desirability bias,” posits that some voters might be unwilling to publicly declare support for Mr. Trump, even though they privately support him. However, Mr. Trump’s support changes little depending on the polling method. Instead, Sec. Clinton’s support increases in live caller polls, which accounts for about three-quarters of the difference between the two methods. Part of this effect may be related to a smaller share of undecided voters in live caller polls, which often ask undecided respondents which way they “lean” to reduce the undecided share. The undecided vote has declined to less than 3% in recent live-caller polls, roughly half the level of automated polling.

While a large third party and undecided vote create more uncertainty, it’s probably not enough to overcome Sec. Clinton’s now-larger polling lead. As often occurs, the share of voters supporting third-party candidates has declined as Election Day draws near; in polling after the last debate, it has declined to an average of 7.5% of the vote, from nearly 12% in August In recent polling, the undecided share has declined as well, and is now around 5%, roughly the same size as Sec. Clinton’s lead in the average of recent polls.

In theory, if undecided voters broke entirely in favor of Mr. Trump on Election Day, this could change the election outcome. However, the views of undecided and third-party voters suggest that they are more likely to vote for Sec. Clinton than Mr. Trump, if they vote at all. A recent Washington Post poll reported that 46% of “likely voters” who support neither Sec. Clinton nor Mr. Trump had a “strongly unfavorable” view of Sec. Clinton, while 71% had a similar view of Mr. Trump. Larger differences existed in views regarding qualification to be president, in favor of Sec. Clinton. Moreover, as noted above, the polls that push undecided voters to choose a candidate show Sec. Clinton with a larger lead, suggesting that the majority of undecided voters prefer her over Mr. Trump.