by Stephen Soukup
Polling depends heavily on the pollster’s “assumptions,” the choices he or she makes about the variables that will be included and the weightings to give those variables.
For decades, every two years, on the morning of the first Tuesday of November, I would make an official, professional prediction about the likely outcome of the elections. That was my job—or at least part of it—as a Washington analyst for the financial services industry. It says something terribly unfortunate about our society that institutional investors have a profound vested interest in who controls the levers of power in Washington, but that’s just the way it is, and that’s a story for another day.
In general, I was pretty good at calling elections, even when I made difficult or contrarian calls. I knew George W. Bush would win reelection in 2004, for example, even after pollsters like John Zogby had concluded that the race was “John Kerry’s to lose.” If I recall correctly, I called every state that year and every Senate race, many of which challenged conventional wisdom. Likewise, in 2016, I predicted Donald Trump’s victory over Hillary Clinton.
As a rule, the outsider’s key to understanding political polling is remembering that it’s not a science but an art. Yes, there’s a great deal of math involved. And statistics. And all of that requires a certain amount of specialized knowledge and skill. Still, when it comes down to it, polling depends heavily on the pollster’s “assumptions,” the choices he or she makes about the variables that will be included and the weightings to give those variables. Good pollsters don’t possess better math skills or some super-secret specialized knowledge their competitors don’t have. They are better at intuiting which variables matter and at keeping their personal biases from too heavily affecting their weighting decisions.
In the cases of Bush and Trump, many pollsters—being part of the broader political class—wanted them to lose. They wanted Kerry and especially Hillary to win. And in preposterous but entirely comprehensible fashion, they projected their biases onto the electorate. They assumed—spectacularly wrongly—that the voters, in general, shared their prejudices and their preferences. The results spoke for themselves.
To be clear, when this happens, it isn’t because the pollsters are intentionally trying to manipulate voters or “push” the vote in one direction. They are simply “mirror imaging,” which is to say, presuming that their subjects couldn’t possibly think differently about specific topics than they do themselves. There’s nothing nefarious about this, but it is one of the biggest and most significant shortcomings in political polling, exacerbated by the fact that pollsters generally don’t like to discuss or even acknowledge it. For the most part, they’d rather the public see them as oracles of sorts, people who can see the future based solely on the skills they possess as members of an exclusive and enigmatic guild. That’s not true, of course. They’re bias- and mistake-prone just like everyone else, but they would prefer, as a rule, to hide all that behind math and statistics to make their craft seem scientific.
Given all of this, it might appear perfectly reasonable to question the pollsters again this year and to assume that they have everything mixed up again because of their biases. And that may, indeed, be the case. Unfortunately, it’s not obviously the case as it was in 2004 and 2016. The other day, I had a retired client email me and ask if I had any sense of who would win the Senate, the Republicans or the Democrats. I honestly have no idea and, as a result, didn’t reply (sorry, Mike). Yes, the polls look absurdly biased. It doesn’t seem possible, for example, that James Talarico is not only leading the race in Texas but also leading it by a wide margin. Likewise, Dan Osborn in Nebraska and Adam Hamilton in Kansas. None of it makes sense, President Trump’s poll numbers be damned. Even so, I have no idea what to think, largely for three reasons.
First, “early voting”—or perpetual voting, as it should more aptly be called—has altered the ways campaigns affect elections and the way election polling should be interpreted. Once upon a time, campaigns were built to crescendo the weekend before Election Day. Candidates would drive their message home, over and over, and by the time voters headed to the polls, the “coming home” effect would swing independents and lukewarm partisans in their direction and would make the idea that they were even in trouble look ridiculous, in retrospect. Similarly, campaigns planned surprise revelations for the last moment. They call them “October surprises,” after all, not July surprises or August surprises. They’re last-minute shock bombs.
None of this makes much of a difference, though, when there is no Election “Day,” when voters can go to the polls for three full weeks. In the old days—i.e., before COVID prompted widespread no-excuse early voting—this weekend’s endorsement by Democrat Haley Stevens of Republican Senate nominee Mike Rogers probably would have flipped the race to Rogers. But now? Who the heck knows? Maybe it will. Or maybe those voters who might have been persuaded by that endorsement cast their ballots two weeks ago last Thursday.
A second reason that polling and its clear shortcomings are difficult to read right now is the rise of betting markets like Kalshi and Polymarket. As a believer in the predictive power of markets, I should, by all rights, think that market-based projections are better, less susceptible to bias, and more reliable than polls. And since all (or most) of the high-profile contests are currently yielding the same results as the polls, I should assume that the polls are right. Right? Wisdom of crowds, and all that? Except that I don’t make any such assumption. In forecasting election outcomes, betting markets are sometimes quite valuable, sometimes marginally helpful, and most of the time simply redundant. In 2024, for example, betting markets had Trump winning, defying the polls—and they were right! That’s valuable information and an important precedent. But that doesn’t mean that they’re right this time.
For starters, the biggest problem with betting markets is that sometimes, their main input on elections is… polling averages. If bettors are using polling data to make their wagers, then the markets add no independent signal. It’s only one signal, counted twice and presumed, therefore, to be confirmatory, when it’s actually no such thing. As Robert Gilhooly, the senior emerging markets economist at Aberdeen Investments, noted two weeks ago, compared to presidential elections (like 2024), Senate contests tend to be “thin markets,” meaning that they have “fewer traders and less independent information.” In turn, this means that polls become a more significant input and they “shape traders’ expectations.” Thin-market bets offer very little by way of added predictive value.
A third reason that the polling in this year’s election cycle offers so few substantive clues about the likely outcome is the bizarre and almost pathological behavior of the political class. As I have noted before in these pages, my personal favorite expert on conspiracy theories is the inimitable Daniel Pipes, the president of the Middle East Forum, who has written two excellent books on the subject and draws a distinction between conspiracies, conspiracy theories, and what he calls conspiracism.
“Conspiracies,” Pipes notes, are real, and they have existed throughout history…. These are facts. “Conspiracy theories,” by contrast, are a mixed bag. They propose a hypothesis—an attempt to explain events by positing a hidden coordinating group. Most such theories are wrong, but some turn out to be correct. Others, while wrong, are arrived at through reasonable inference from available evidence. “Conspiracism,” on the other hand, is simply pathological. It is…a “worldview,” a standing assumption that conspiracy is the normal mechanism by which history moves, that powerful hidden forces are always the real explanation behind surface events.
Ever since COVID, which demonstrated that many of our formerly most trusted institutions are unworthy of the faith we once placed in them, conspiracism among the general public has surged appreciably—and understandably. If we can’t trust our government, our educational establishment, our medical establishment, and the media not to lie to us about a global pandemic, then on what matters can we trust them?
The behavior of many of those same institutions during this campaign season seems almost designed to make the COVID betrayals pale in comparison. In the past few weeks, several erstwhile conspiracy theories have been proven to be—or nearly proven to be—genuine conspiracies.
For years, conservatives in Minnesota and elsewhere have insisted that Somali-American Congresswoman Ilhan Omar had committed immigration fraud and had likely married her brother as part of the scheme. They were called inveterate racists for their trouble. Now—thanks in part to the Trump administration’s much-maligned raid on the Congresswoman’s home—we know that there is evidence that she was involved in immigration fraud of some sort (perhaps on her father’s part) and that even the Obama Justice Department found evidence that she “may have entered into a fraudulent marriage.”
For weeks, President Trump and his administration have been arguing that the Senate race in Alaska has been tainted, perhaps criminally, by the presence of a second man named Dan Sullivan, running as a Republican, against the incumbent Republican Senator Dan Sullivan. Democrats called this “sour grapes” from the GOP, at best, and, at worst, an attempt to invalidate a fair election before it even took place. Last week, we learned that the whole thing was, indeed, an orchestrated campaign and that the Democratic Party, including the campaign of Democratic candidate Mary Peltola, “contributed to” the scheme. As The New York Times put it:
The Peltola campaign has repeatedly denied any involvement in [the second Dan Sullivan’s] candidacy. But The Times has found, based on interviews and text messages, along with documents prepared for a potential lawsuit against the campaign, that not only did Ms. Peltola’s team work with people trying to recruit Mr. Sullivan to run for Senate, but the effort was much broader. The group worked to coax other would-be spoilers into the race, to dilute the incumbent’s support and help catapult the Democrat to victory.
Forever (or so it seems), Republicans have insisted that some political polls are phony, that they are made up, that they’re fakes intended to push or dishearten voters. Needless to say, these claims have almost always been dismissed as the rantings of disgruntled nutters. And yet… Twice this campaign season, polling operations have been shown to be faking their polls. In August, a group called Median Strategies shut down after it was shown to be fabricating polls. Charles Franklin, director of the Marquette Law School Poll, said of the operation, “it’s not a poll; it’s just a fraud.” Then, more dramatically, in October, Wedgewood Polls admitted to faking its data, claiming that it was conducting an “experimental project to demonstrate the vulnerabilities, lack of oversight, and overall low barrier to entry in the modern political polling industry.” Polling veteran Nate Silver called that explanation “bullsh-t,” and noted there was “a pattern here of Wedgewood showing better numbers for notably progressive candidates than the consensus.” The biggest problem with Wedgewood was that it “covered” multiple competitive races (including Michigan, Maine, Mississippi, Nebraska, and Texas), and its numbers were incorporated into aggregators like The New York Times, Decision Desk HQ, and others. Relevantly, Silver also argued that “One of [the phony pollsters] is publishing fake polls to manipulate prediction markets.” Fancy that.
Evidence from previous elections has shown that polls have been less reliable in recent years, in part because fewer people are responding. Moreover, fewer people are responding, in part, because they don’t trust polls or those who conduct them. It’s not hard to see why.
Taken individually, each of these conspiracy-theories-turned-conspiracies would be evidence of malfeasance on the part of members of the political class, often members of the government itself. Taken together, they are justification for the ongoing collapse of Americans’ trust in that political class and the institutions they supposedly defend. If I were a conspiracist myself, I might be inclined to think it was all part of a larger plot by more established figures in the political class to sow distrust in and rejection of those institutions, in preparation for a coordinated post-midterm legitimacy crisis that would trigger mass protests and noncooperation, force President Trump to invoke the Insurrection Act, and use the resulting chaos to rewrite or replace key elements of the U.S. constitutional order.
Of course, I’m not a conspiracist, so I don’t think that. At the same time, however, I can understand why people would. All of this—from COVID to today—has legitimately undermined trust in the institutions, which, almost definitionally, sparked a significant and ongoing rise in conspiracism among the general public. On the Right, the culprit is the government itself and its allies in the private and nonprofit sectors. On the Left, the culprit is the eternal culprit, the group that gets blamed for everything, everywhere, at all times, and which, according to Pipes, constitutes one of the two major antagonists in all appearances of conspiracism: the Jewish people. To conspiracism’s adherents, the conspiracies of the last few years serve as vindication, a sign that bigger, even more nefarious plots are afoot.
In the end, all of this makes society more contentious, more divided, and less effectively governable. It also makes elections harder to forecast. I think—based on nothing more than historical patterns—that today will be a good day for the Democrats. But I wouldn’t be surprised by anything
Photo: Close-up of "Vote Here" directional sign outside of polling place
Stephen Soukup is the Director of The Political Forum Institute and the author of The Dictatorship of Woke Capital (Encounter, 2021, 2023)
Source: https://amgreatness.com/2026/11/03/predictions-are-hard-especially-about-the-future/
No comments:
Post a Comment