The Midterm Polls Are LYING

Pollsters think they know who's about to win the midterms. But here's why you shouldn't trust any of them. Welcome to American Covered. I'm Chris Chappell. With a week to go before the midterms, you've probably been hit with a flood of articles, videos, and social media posts telling you which party is going to come out on top. And people are betting Little Timmy's College Fund on Polymarket based on these posts. But I'm here to tell you that regardless of your political affiliation, You shouldn't listen to any of these people. The media, the online markets, they're useless. Actually, they're less than useless. First, a little context on how Americans vote. Or rather, how they don't vote. For people who take their rights pretty seriously, Americans are pretty bad at showing up to the polls. Take 2020. It had the highest voter turnout rate since the early 1900s, with the supercharged political atmosphere and the proliferation of mail-in ballots. But even then, American turnout was only 62%. That's still below average for a developed country. And that was a presidential year. Midterms are way worse. Since 2014, turnout for those races have gone from as low as 34% to a high of 48%, which is to say, less than half of eligible voters. Now, you may have heard that millions of Americans have already cast their votes, and that early voting is way up this year. And that's true. But early voters aren't really representative of the broader voting public. As Axios writes, those casting ballots ahead of Election Day tend to be more politically engaged. Basically, the people who vote early were always going to vote. The question is about everyone else. And in a midterm election, everyone else is a lot of people. So if early votes aren't the best metric to go by, what can we use? Well, polling, of course. But opinion polls have a few glaring problems. First of all, almost nobody answers them anymore. Pew, one of the most respected pollsters in the country, has watched its phone response rates collapse to single digits. That means for every 100 people a pollster calls, about 94 hang up or let it go to voicemail. As a result, a lot of pollsters have moved to online panels and text surveys to try and make up for it. But that doesn't fix the second problem. The people who do respond aren't like most people. Chances are they're more engaged than the average voter, more civic-minded, if you will. In one Pew study, 55% of survey respondents said they did volunteer work in the past year, compared to around 27% of the general population. And then there's the opposite problem. People who distrust institutions tend to skip surveys entirely. After 2020, when the polls had their worst performance in 40 years, much of the blame was placed on the sheer number of non-responses, which could pose an even greater crisis for political polling than the advent of mobile phones. Now, pollsters can try and weigh the results to account for all the people who didn't respond, but Pew Research admits that it may not fully correct the bias. At the end of the day, weighting boils down to a judgment call on the part of the pollster. Pollsters have to decide who's likely to vote, how much each group counts, and what to do with the undecideds. Case in point, in 2016, the New York Times ran an experiment where they gave four different pollsters the exact same raw poll data and asked them to calculate the result. They got back four different results spread across five points. So the data is shaky at best, but it gets worse once you look at the incentives of the people collecting it. Imagine your pollster. If you go along with what all the other pollsters are saying and it turns out you're all wrong, well, that's fine. We all make mistakes. But if you break with the consensus and your predictions don't pan out, well, then you're singled out as a laughingstock. Just ask Ann Seltzer. For years, she was considered the gold standard in Iowa's polling. Then, days before the 2024 election, she published a poll showing Kamala Harris up three points in a state where Trump was considered a safe bet. And they were right. Iowa was a safe bet for Trump. He won the state by 13 points. Now, this was Seltzer's last poll before retiring from election polling. But it was still a pretty embarrassing note to go out on, and the headlines reflected that. The same thing happened to Gallup in 2012. Their final poll for that year's presidential election had Romney narrowly ahead. And then Obama won by about four points. Gallup soon gave up on trying to predict presidential outcomes, and sat out 2016 entirely. That was probably for the best. God knows how many pollsters had their predictions blow up in their faces that year. But say you want to keep on polling, but you don't want to look like an idiot when your predictions turn out wrong. What do you do? Well, you stick close to what other pollsters are saying. In the industry, this is called herding. In 2024, Nate Silver ran the numbers on the swing state polls and found they were clustered around a tie far more tightly than random sampling should allow. He said that this is a clear-as-day example of what we call herding— the tendency of some polling firms to move with the flock by not publishing results that don't match the consensus or torturing their turnout models until they do. You also need to ask the question of who exactly is paying for these polls. Because a good chunk of public polls are sponsored by news outlets. And Race Tightens in the Final Stretch gets a lot more clicks than Nothing Has Really Changed. And don't forget, a lot of polls come from the political campaigns themselves. Last year, polls in the Texas Senate primary showed incumbent John Cornyn closing the gap on Ken Paxton. One of them came from the Senate Leadership Fund, a super PAC backing Cornyn. And how'd that turn out? Cornyn lost the runoff by more than 27 points. So if the polls aren't all that useful, what about the betting markets? Are they any more trustworthy? Well, here's something you might not know. A lot of the networks reporting on the odds for any given race are in business with the companies setting them. CNN and CNBC have deals with Kalshi. Dow Jones, the Wall Street Journal's parent company, has partnered with Polymarket. Here's the problem with that. A poll typically gives every respondent one answer. A betting market typically gives the most influence to whoever bets the most money. And that means a few big players known as whales can move the odds all by themselves. In 2024, a French trader who called himself Theo bet $30 million on Trump through four polymarket accounts. His bets visibly pushed up Trump's odds. At one point, his accounts held about 25% of the contracts on Trump winning the electoral college. Now, Theo turned out to be right and he made a fortune. But that's not the point. The point is that one guy with a big enough bankroll could move the number millions of people were watching. And that's a national presidential race with billions of dollars in bets. Smaller races are much easier to push around. As of last week, the entire prediction market for Ohio Governor had about 2.8 million traded. That's chump change compared to what campaigns spend on TV ads. And here's where betting markets get even weirder. When a Polymarket bet is disputed, the outcome gets decided by a vote among holders of an independent cryptocurrency called UMA. And Bloomberg found that vote is dominated by a handful of anonymous wallets. So whales can move the odds before the event. And a different set of whales get to decide what officially happened after. And a lot of that money may not even be real. Columbia University researchers estimated that in December of 2024, around 60% of Polymarket's trading volume was wash trading, people buying and selling to themselves to fake activity. Why does this matter? Because those numbers wind up influencing the public. After 2016, researchers tested what election forecasts do to voters, and they found that forecasting increases certainty about an election's outcome, confuses many, and decreases turnout. out. When many people see a candidate with high odds of winning, they treat the race as already decided. When a race feels decided, well, what's the point of voting? It's basically a psy-op. Not in the sense that there's people in some smoky backroom thinking up ways to suppress voters, but in the sense that polling and predictions are an industry, an industry with certain incentives built into them, managed by people using flawed methodologies. So what should we take from all this? The polls are built on a tiny sliver of people, fed through a system built to protect the pollster from personal embarrassment. The betting markets are run by whales, padded with fake volume, and broadcast by the networks that partner with them. Neither of them can tell you who's going to win a close race. Meanwhile, in a country where about half of eligible voters skip the midterms, elections get decided by whoever bothers to show up. In a close race, you are the margin of error. So don't let a poll, a betting market, or a headline decide whether you vote. Don't take my word on the rules, either. Go to vote.gov, find your state, and make a plan to go vote. But what do you think? Leave your comment below. I do read as many as I can. But I read every comment on AmericaUncovered.tv, where you'll get access to thousands of episodes ad-free, plus a community chat, live streams, and more. Just click this button here or the link below. Once again, I'm Chris Chappell. Thank you for watching America Uncovered.