Election Day Upsets Are Still Possible

W hen projecting election outcomes on the basis of public polling, there are two questions worth asking: What is probable, and what is possible? If you always bet that the candidate leading in the polls will win, you’ll be right way more often than you’ll be wrong. A more sophisticated understanding, though, will take into account the average error across poll averages as well as the biggest wild cards: the leader’s margin and share of the vote, the number of undecideds in the poll, how many polls of the race have been taken recently, and the track record of pollsters and polls in a particular state.

That approach will give you a good idea of what’s probable. If you want to know what’s possible, on the other hand, you should start at the end rather than the beginning, isolating past races in which the polls got the outcome wrong. These are not the races that offer us the most likely parallels to the typical 2022 election; they are the outliers. Nor are they necessarily the races in which the gap between the polls and the results was the biggest; every year, in a certain number of elections, polls miss the margins of victory by a lot while still getting the winners right. Studying only the candidates who outperformed their polling numbers is a classic example of survivorship bias, which is problematic if you’re projecting probabilities. But it is still useful for setting candidates’ and parties’ ceilings and floors heading into Election Day.

From 2010 to today, there have been three midterms and three presidential elections. Democrats have won two of the presidential races to Republicans’ one, and Republicans have had two midterm wave elections to Democrats’ one. Overall in that period, there have been 522 regularly scheduled statewide elections for president, senator, or governor (counting D.C.’s presidential votes but not counting recall elections, special Senate elections, or runoffs): 210 statewide elections for Senate seats, 159 statewide elections for governorships, and 153 statewide elections for president.

Among those, there were 34 in which the winner did not lead in the final RealClearPolitics poll average: 33 that were won by the candidate who trailed, and one — the 2018 Wisconsin governor’s race, won by Tony Evers — in which the final RCP average was a tie. (I counted David Perdue’s getting more votes than Jon Ossoff on Election Day in 2020 as an upset, even though Perdue lost to Ossoff in the subsequent runoff election.) Fourteen of those were Senate races, 14 were governor’s races, and six were presidential races, which are more extensively polled in competitive states. Thirty-four misses may sound like a lot, but even excluding the 2018 Wisconsin governor’s race, that averages out to a polling-miss rate of one out of every 12.2 elections for governor, one out of every 15 Senate elections, and one out of every 25.5 presidential elections. If those averages hold, we could expect two Senate races and three races for governor to be won by candidates who go into tomorrow’s elections still behind in the polls.

Let’s take a look at what we can generalize about the races the polls got wrong:

• Twenty-one of them missed Republican victories; 13 missed Democratic victories.
• Eight of the 13 Democratic victories the polls missed were in 2010 or 2012, and three were in the 2018 Democratic wave. By contrast, 14 of the 21 Republican wins missed by the polls were between 2016 and 2020 (six of the other seven were in the Republican wave year of 2014, and the seventh was in the following year’s race for governor of Kentucky). That suggests both that the tendency to underestimate Republicans is growing, and that it is more pronounced in good Republican years.
• Most of these polling misses came in competitive states rather than deep-red or deep-blue states, which is not terribly surprising.
• Several states have been the site of repeated polling failures: Florida five times (four of them Republican wins), Wisconsin three times (two of them Republican wins), Kansas three times (two of them Republican wins), and Illinois, Montana, North Carolina, Nevada, and Pennsylvania twice each.
• In 2018 in Florida, 2016 in Wisconsin and Pennsylvania, 2014 in Kansas, and 2012 in Montana, the polls missed two races in the same state, in each case in the same direction, suggesting a systemic failure to poll the right voters and project turnout.

There are some further conclusions we can draw from looking at the numbers:

First, in none of these races was the ultimate loser polling at or above 50 percent heading into Election Day. The 50 percent rule of thumb is thus a good one for figuring out which leads are safe. The highest-polling Democratic failures were Andrew Gillum at 49.4 percent in the 2018 Florida governor’s race and Bill Nelson at 48.8 percent, in the 2018 Florida Senate race. The highest-polling Republican failures were Mitt Romney at 49.7 percent in Florida in 2012, Rick Berg at 49 percent in the North Dakota Senate race in 2012, Ken Buck at 49.3 percent in the Colorado Senate race in 2010, and Sharron Angle in the Nevada Senate race in 2010.

The average Republican who came from behind trailed 45.9 percent to 43.4 percent, and won 50 percent to 46.6 percent; the average Democrat who came from behind trailed 47.1 percent to 45.1 percent, and won 48.8 percent to 47.1 percent. That tells us a few things: 1) Republicans are particularly prone to staging upsets in races where the polls show a large undecided or third-party bloc all the way to the end; 2) When the polls miss, it’s more often because they underestimated the ultimate winner’s support than because they overestimated the ultimate loser’s; and 3) the misses tend to be larger in Republican upset wins (an average Democratic polling lead of 2.5 points) than in Democratic upset wins (an average Republican polling lead of two points).

The biggest upsets were eleven races in which one side led by at least three points, and in five of those one side led by at least 4.7 points. In those five races, Donald Trump came from 6.5 points back to beat Hillary Clinton in Wisconsin (2016), Heidi Heitkamp came from down 5.7 points to win in North Dakota (2012), Susan Collins came from down five to win by 8.5 points in Maine (2020), Mike DeWine came from down 4.7 to win in Ohio (2018), and Pat Quinn came from down 4.7 to win in Illinois (2010). Overall, the five biggest misses compared with the final outcome were all in races won by Republicans: Collins, Matt Bevin in 2015, Sam Brownback in 2014, Eric Holcomb in 2016, and DeWine, all of whom outperformed their margin in the final polling average by nine or more points.

So, what does all of this mean for tomorrow?

Well, it is certainly possible that those Republicans trailing by six points or less against opponents polling under 50 percent in the RCP average could stage upsets. At this writing, in the Senate, those candidates would be Don Bolduc in New Hampshire (down one, with Maggie Hassan at 48.4 percent), Blake Masters in Arizona (down one, with Mark Kelly at 48.2 percent), Tiffany Smiley in Washington (down 4.3 points, with Patty Murray at 49.3 percent), and Joe O’Dea in Colorado (down 5.3 points, with Michael Bennet at 49.3 percent). (Smiley and O’Dea, as the two trailing their opponents by the most, would be the most surprising upsets.) A few other Republicans — Dr. Oz, Herschel Walker, and Adam Laxalt — are clinging to small leads in races that are considered toss-ups.

In the races for governor, aside from Christine Drazan (who is presently tied with her Democratic opponent in Oregon), the possible Republican upsets would be Derek Schmidt in Kansas (down three points, with Laura Kelly at 46 percent), Mark Ronchetti in New Mexico (down four, with Michelle Lujan Grisham at 48 percent), and Scott Jensen in Minnesota (down 4.3 points, with Tim Walz at 48 percent). It would be much more surprising to see Tudor Dixon win in Michigan (down 4.2 points to Gretchen Whitmer) or Lee Zeldin win in New York (down 6.2 points), because the Democratic nominees in their races entered today at or above 50 percent in the RCP average. A few Republicans — Tim Michels, Joe Lombardo, Kevin Stitt, and Kari Lake — also hold small leads in governor’s races.

The bottom line? Don’t be surprised at some upsets on Tuesday night, especially in favor of Republicans — but remember that those are the exception, not the rule.