With Super Tuesday behind us, it seems like Americans are adjusting to the reality that Donald J. Trump could very well become their next president. How can you tell? For starters, Google searches about moving to Canada exploded to 10-year highs after the pumpkin-hued nativist’s dominant performance during this week’s primaries—a now-traditional sign that U.S. liberals are performatively grappling with dread over their country’s fate.
It’s still premature for Democrats to give into despair. (Not for Republicans, though. Poor Chris Christie.) Trump may be ascendant, but it is still unclear he can win a general election. Both Hillary Clinton and Bernie Sanders lead Trump in polling averages. Trump is also deeply disliked by voters—moreso, even, than Clinton herself, which is a feat. And yet the surveys showing a Democratic edge over the Donald are only cold comfort, since polls at this stage of a campaign still aren’t very accurate at predicting the outcome in November.
When will they become more accurate? Political scientists Robert Erikson, of Columbia University, and Christopher Wlezien, of the University of Texas at Austin, looked into this question for their book The Timeline of Presidential Elections: How Campaigns Do (and Do Not) Matter, which they updated as a short and handy e-book after the 2012 race. “Polls from the beginning of the election year reveal almost nothing about the final vote,” they write. “Preferences start to come into focus by mid-April, when candidates have been identified and matched in pollsters’ trial heats.” But many voters seem to make up their minds for good after the party conventions in the summer. “Once the conventions are over, further campaign events—even presidential debates—rarely result in dramatic change.”
Here’s how that looks graphed out. The left axis is essentially a measure of how predictive presidential polls are at each stage of the campaign: A score of 1 would mean that polls perfectly anticipate the winner, while a score of 0 would mean they have no forecasting power at all. We’re currently 251 days from the election, meaning polls are only somewhat suggestive of what the final outcome might be. They become a more accurate gauge later in the spring, but things don’t really settle until after the conventions.
Here’s another way of visualizing the pattern. In the graphs below, the closer the dot plot lines up in a straight, 45 degree angle, the closer the polls come to predicting the actual democratic share of votes on Election Day. As you can see, even by April, things are still a bit scattered. The picture clarifies after the conventions give one candidate a more permanent bounce.
Why are polls so inaccurate at this stage? The simple, obvious answer is that people are still making up their minds. The more nuanced version, though, is that voters get more partisan as the election rolls on. “In April, with the election a distant seven months in the future, survey respondents do not strongly adhere to their party identifications when asked about presidential candidates,” Erikson and Wlezien write. “In effect, in April, partisan voters behave more like Independents than they do on Election Day. Like Independents, April poll respondents in the aggregate tilt one-sidedly to the candidate most in favor at the moment.”
In other words, early on, voters like whoever the hot candidate of the moment is. As the campaign unfolds, though, people start settling into their partisan camps as they pay more attention to candidates’ ideologies and how they matches their own inclinations. The conventions act as giant, one-week commercials for what each party actually stands for, which helps to clear up the differences between them.
So here are my takeaways. First, you should probably be cautious about reading too much into polls that suggest outsider, unorthodox candidates like Trump or Bernie Sanders have a great degree of crossover appeal. It’s possible they do—I don’t personally find it likely, but maybe Trump truly is about to assemble an entirely new coalition of voters, in which case, old partisan divides might not matter much. But it’s also likely that Americans haven’t thought very hard—or at all—about the candidates’ policy positions (not that Trump has given them a ton to think about), and are reacting to vague impressions from positive (or, in Trump’s case, ubiquitous) press coverage.
Second, if you feel a wave of nausea at the thought of Donald Trump gold-plating the White House and he’s in the lead come April, you might want to take a vacation to Vancouver or Montreal just to get a feel for life up north. If he’s leading after the convention, then it’s time to look at real estate listings. You know, just in case you want to get in before the post-election rush.