As you probably know if you’ve stepped anywhere near the news in the past few weeks, this presidential race is the closest in many a year. What the news hasn’t necessarily emphasized is just how close the polls show it being — and how a slight error in modeling or a last-minute shift in patterns could mean that either candidate ends up walking away with an Electoral College landslide.
As of Friday, there are nine toss-up states, according to the RealClearPolitics polling aggregate: Arizona, Georgia, Michigan, Minnesota, Nevada, New Hampshire, North Carolina, Pennsylvania and Wisconsin. Add in Nebraska’s 2nd Congressional District — Nebraska apportions some of its electoral vote total by the winner in each district — and that’s 108 electoral votes that are toss-ups as the final polls begin to roll in.
In five of those states — Arizona, Georgia, Nevada, North Carolina and Pennsylvania — Donald Trump holds a lead; Kamala Harris leads the rest. However, aside from Minnesota, which is just within the toss-up parameters but which leans Harris — none of those leads is larger than 2.5 points in the aggregate.
And when you consider how much the polls were off by in the past, that’s an issue.
How big of an issue? Consider that ABC News looked at it with data from its own sister site, wonk outlet FiveThirtyEight, which only had seven swing states. (Minnesota and New Hampshire were off the table.) If the same polling errors that overestimated Joe Biden’s support against Donald Trump apply again four years later, Trump would be walking away with 312 electoral votes.
“Of course, if the polls are off, it won’t necessarily benefit Trump,” ABC News noted in the Wednesday report.
“The direction of polling error is impossible to predict in advance, and polls have overestimated Republicans plenty of times in the past. In a scenario where the polls overestimate Trump’s margin by 4 points in every state, Harris would win all seven swing states and 319 electoral votes.”
And while the data from FiveThirtyEight’s data model shows that the polling error will be roughly 3.8 points, that error “is not uniform across states,” but “generally speaking, when polls overestimate a candidate, they tend to overestimate them across the board.”
Perhaps more alarming, for those of you looking to take solace in a 2016-like and 2020-like error in polling giving this to the GOP automatically: “In 50 percent of the model’s simulations, Trump beats his polls, and 50 percent of the time, Harris does.”
“Given that all seven key swing states are so close, even small polling errors in the same direction can have a big impact on who wins the election. According to the simulations from our model, there is a 60-in-100 chance either candidate wins over 300 Electoral College votes — which Harris could do by winning five of the seven swing states and Trump six out of the seven.”
Furthermore, with the closeness of the polls, it might be important to note what you’re actually reading when you read both a national poll and a statewide poll.
First off, national polls: While they give you a barometer on how the country views a candidate, the popular vote means nothing. Winning the popular vote for the presidency and $5.69 will buy you a Big Mac and nothing else.
However, popular vote polls can sort of give you an idea of how a candidate is doing. In both 2016 and 2020, Trump outperformed the polling aggregates on a national level substantially.
The difference was less marked in 2016, where Hillary Clinton had a 3.2 percent lead in the national RealClearPolitics aggregate only to win the popular vote by 2.1 percent. However, the problem there wasn’t how wrong the national polls were but how wrong they were on a swing state level — particularly in the so-called “Blue Wall” states of the Midwest.
In 2020, the difference was much starker. Biden led by 7.2 percent in the final RealClearPolitics aggregate but only won by 4.5 percent — and the results weren’t known until well after election night.
If either of those errors repeat themselves, Harris is in trouble.
Trump actually leads in the aggregate as of Friday morning by a slim margin, 48.5 to 48.0 percent. The general thought is that any Democratic candidate should have at least a few points in hand in the national vote, since — thanks to the fact that the GOP has limited reason to contest the election in many of America’s most populous cities or states — Harris should be up more times than she’s not in polls if she’s to have any chance of winning.
And keep in mind that if she does worse in the swing states, like Hillary Clinton did, that’s still problematic. As of right now, if Trump merely wins the states he’s leading in as of the RealClearPolitics aggregate Friday, the GOP will win.
However, it could be even worse than that: In Michigan, Harris leads by just 0.4 percent, and in Wisconsin, 0.2 percent. Consider how Biden’s vote went there vs. the aggregate in 2020. He was up by 7.3 in Michigan but only won by 2.8. In Wisconsin, he led by 5.7 percent in the final aggregate, but only won by 0.7 percent.
Finally, it’s worth noting how polls work. I know we all think we do, but it turns out some people can be a bit foggy on it.
First, this explanation from the Pew Research Center: “Because surveys only talk to a sample of the population, we know that the result probably won’t exactly match the ‘true’ result that we would get if we interviewed everyone in the population. The margin of sampling error describes how close we can reasonably expect a survey result to fall relative to the true population value. A margin of error of plus or minus 3 percentage points at the 95 percent confidence level means that if we fielded the same survey 100 times, we would expect the result to be within 3 percentage points of the true population value 95 of those times.”
However, when you see 3 percent margin of error, people forget that doesn’t just mean it would only be off by 3 percentage points, full stop, 95 percent of times that they run it. In fact, as The New York Times pointed out one month prior to the 2016 election, you should take that number and double it:
If 54 percent of people support Hillary Clinton, the survey estimate might be as high as 57 percent or as low as 51 percent, but it is unlikely to be 49 percent. This truism of modern polling, heralded as one of the great success stories of statistics, is included in textbooks and taught in college classes, including our own.
But the real-world margin of error of election polls is not three percentage points. It is about twice as big.
In a new paper with Andrew Gelman and Houshmand Shirani-Mehr, we examined 4,221 late-campaign polls — every public poll we could find — for 608 state-level presidential, Senate and governor’s races between 1998 and 2014. Comparing those polls’ results with actual electoral results, we find the historical margin of error is plus or minus six to seven percentage points. (Yes, that’s an error range of 12 to 14 points, not the typically reported 6 or 7.)
That being said, aggregates do take a little bit out of that equation by reporting on numerous polls and weighting them various ways before they’re averaged. But what happens if the errors are vast and tend to go in one direction, as FiveThirtyEight says its models tend to show?
The point is: This race is far from over. The numbers don’t lie in that respect, and every vote counts. With just a few days left in this long, curious election cycle, it’s still too early for anyone to say either candidate has anything resembling a clear-cut advantage.
Advertise with The Western Journal and reach millions of highly engaged readers, while supporting our work. Advertise Today.