On Thursday, four polls of the Georgia Senate race between Republican challenger Herschel Walker and Democratic Sen. Raphael Warnock were released. One showed Warnock leading by 5 points. Another had him leading by 2. A third had Walker leading by 1. And another had Walker leading by 4.
These were all reputable pollsters with grades of B or above in FiveThirtyEight’s ratings. Their methodologies seemed fine. They simply don’t know exactly what’s going to happen in a close race and are—each of them—making their best guesses with imprecise tools.
That’s the story everywhere, really. It’s not just Georgia. Nate Cohn is the longtime editor of the New York Times’ Upshot section; he’s seen as much midterm poll data as anyone in the country and probably knows as much as anyone about the statistical analysis that goes into making it. He wrote an Oct. 25 piece called “If These Poll Results Keep Up, Expect Anything on Election Night.” On Monday, the Economist’s data journalist G. Elliott Morris published his own … prediction?
The data expert consensus is essentially “shruggie emoji.”
The major parties think they might know what’s going to happen, and it’s possible to get a sense of what their internal polling is telling them by looking at where they send TV money in the last few weeks of a race. (Right now that’s “districts where Democrats thought they were going to be safe but maybe aren’t.”) But “internal” doesn’t mean “magic,” and they can be just as wrong as anyone else—like in 2020, when Republicans did much better nationally than their own secret numbers said they would.
Polls have always been estimates that, by necessity, require guesswork. That job is getting harder, and it’s not even entirely clear why. The American Association for Public Opinion Research published a report in 2021 about why the previous year’s polls were so far off. It noted that “among polls conducted in the final two weeks, the average error on the margin in either direction was 4.5 points for national popular vote polls and 5.1 points for state-level presidential polls.” Five points—during two weeks when almost everyone had made up their minds and many were already voting.
The report concluded, “Identifying conclusively why polls overstated the Democratic-Republican margin relative to the certified vote appears to be impossible with the available data.” Impossible! It outlines what might have happened, namely that right-wing distrust toward polls (and toward institutions in general) may have contributed to a disproportionate number of Trump supporters choosing not to participate in polls by ignoring calls/texts/etc. from pollsters. But to know if that’s the case for sure, one would have to collect data on these nonrespondents. Which would require that they … respond to a poll.
So, we should expect anything—and we don’t know why, exactly, we have to do that. This is true not just in the strict sense of how Americans will vote, but why and what for. The past three presidential elections have involved the electorate (and some crucial number of swing voters within it!) pivoting back and forth between the following: first, an even-tempered liberal whose victory was a historic civil rights event; next, a right-wing TV performer who became involved in politics by circulating false and racially motivated claims about the aforementioned liberal; then, that liberal’s generic, moderate former vice president.
In between, the party controlling the White House has lost every midterm election very, very badly. Meanwhile, of everything that has happened in U.S. politics since 9/11, the Jan. 6 riot was probably the most universally acknowledged at the time, across parties, to have been bad. And what happened to the Republican leaders who were blamed for it by a majority of members of the public? They, ah … they continued to hold power in their party and could win control of both chambers of Congress this week.
“Thermostatic” public opinion isn’t new, and it’s always been silly to draw zeitgeist conclusions about what “people” in the United States want from election results. Distilling 200 million–plus eligible voters’ worth of wildly diverging ideologies, weird personal idiosyncrasies, and conflicting impulses into a sweeping statement about outcomes being a “rebuke” of one abstraction and a “triumph” of another is an ambitious project.
But it is now, empirically, a sillier practice than ever, with the electorate being increasingly more polarized and harder to parse. The country elected both Nixon and Carter, sure, but it didn’t subsequently elect Nixon again on a promise to create an even more corrupt “Watergate II” scandal. With Donald Trump threatening another presidential run to vindicate his 2020 “Stop the Steal” mob, the analogous outcome is on the table today.
What, again, about this midterm, though? The two analysts most universally acknowledged as experts on their states, Nevada’s Jon Ralston and Iowa’s Ann Selzer, have issued final forecasts that would point, if extrapolated nationally, to Republicans’ probably winning the House but not the Senate. Or, in other words, a slight rearrangement of our current muddled state, disruptive enough to make everyone upset but not to settle anything definitively.
But who’s going to let that stop them from taking a big swing about what it all frickin’ means? Whatever happens Tuesday, the political left, right, and center will all claim it as validation of their prior convictions—just like the Bill of Rights entitles them to. (According to Axios, the center has already started doing it.) This country has run on unprovably sweeping assertions since the beginning, and it is time, once again, to plunge into the widening gyre to do some punditry. Who’s coming with me?