March 2026 saw a slew of elections, as the country ramped up to the November midterms. Among these elections were a pair of exciting Texas primaries: State Representative James Talarico triumphed over his opponent, the firebrand Jasmine Crockett, while the top two Republican candidates — John Cornyn and Ken Paxton — advanced to a runoff.
Despite lacking name recognition, Talarico’s victory didn’t come as a surprise to the steady hands at Polymarket; he had been consistently favored to win for more than three months before the election. The Republican primary was a different case. Ken Paxton had been favored to garner the most votes, if not win outright, by online bookies since November 2025 — until about 7:30 p.m. on election night, when Paxton’s chances dramatically nosedived while Cornyn’s rose in tandem. Within an hour, the forecast for Paxton dropped from 89 percent to 33 percent. Cornyn received the most votes in the first round, and is presently favored to win the runoff at 61 percent.
I was watching the online market that night, fascinated by what I saw: Polymarket had been wrong. Not only had it been wrong, but it had been drastically wrong.
Paxton’s chances in the primary, once high at 80 percent, still haven’t recovered from getting second place, even though Cornyn beat him by just over a single percentage point. Cornyn is now favored to win the runoff, a dramatic change from the previous five months. Watching the markets wildly reverse their predictions on election night, Polymarket’s numerous flaws as a bellwether for political outcomes began to reveal themselves.
Polymarket odds for the winner of the first round of the Texas Primary, Feb. 28-March 7 | Image Source: Polymarket
Polymarket, along with fellow betting platform Kalshi, garnered fame when it correctly predicted Donald Trump’s 2024 victory long before pollsters did. Since then, Polymarket has claimed it possesses over 90 percent predictive accuracy a month out from any event in question, and as a result, has begun to reshape the electoral landscape. Already, news sites are reporting Polymarket odds as a candidate metric during races — seemingly because news sites recognize that Americans are happy to look for a more “certain” alternative to polling.
It seems likely that betting sites will continue to play an even larger role in American politics, as their statistics could potentially start swaying electorates to vote for candidates that have better odds for victory. If voters see that a candidate they would otherwise back is rated as having little chance of victory, then they may decide supporting that candidate is pointless. Since betting app odds are determined partially by the amount of money bet for or against a position, this runs the risk of giving wealthy bettors an outsized voice in affecting public opinion. In the 2026 midterms, where a single election in a battleground state could determine which party controls the Senate, such a voice would echo across American society and through the halls of Congress.
No doubt, the 2026 midterms will be the biggest test yet for the bevvy of betting apps. In the Texas primary, Polymarket already demonstrated that it (and other prediction markets) can be incredibly misguided in their predictions — a reality that may diminish their credibility as predictive resources. Now, Polymarket odds are themselves at odds with the conventional wisdom of pollsters. At the time of writing, Polymarket forecasts a 51 percent chance for Democrats to retake the Senate. Taken in conjunction with a 51 percent chance for Democrats to sweep the Senate and House, and a 27 percent chance (the highest of the given options) that Republicans will hold 47 or fewer seats after the midterms, Polymarket seems to be anticipating a blue wave.
In comparison, most pollsters seem hesitant to predict a Chuck Schumer-led Senate come January 2027. Several polls currently rate 45 seats as ‘safely Democratic,’ while homing in on four toss-up races: Maine, North Carolina, Georgia, and Michigan. Even if Democrats can win all four of these races (and Polymarket thinks they are likely to), they would still be stuck at 49 seats. Winning a majority (to say nothing of the nigh-impossible opinion of Polymarket bettors that the Democrats could achieve 53 or more seats), Dems would have to pull off victories in two states that most traditional polls currently rate as leaning Republican: Alaska, Ohio, or even Texas. Here, too, Polymarket is optimistic. Democrats currently have a 57 percent shot at pulling off the flip in Ohio, while the Alaska race shows Mary Peltola pulling away at 62 percent odds.
So why the disconnect? Prediction markets and polls certainly aren’t severed from one another. Polls can influence prediction markets: after a March 29 poll showed the aforementioned Peltola leading her opponent, incumbent Dan Sullivan, by 5 points, her odds of victory went from 52 percent to 62 percent. Though a seemingly small increase, the impacts of that poll caused the markets to go from virtually tied to showing Peltola with a noticeable lead. It’s possible that voters (who may later be polled) could be influenced by what they see on the prediction markets. It’s not a perfectly circular feedback loop, but the two metrics are undeniably linked — and such a link would imply that some deeper reason must account for the difference between the polls and the markets.
First, prediction markets have to choose. There are no “toss-ups” on Polymarket, at least not in the traditional polling sense. One candidate must win, the other must lose. This binary means that bettors are not just predicting the current mood of voters — they must also account for undecided voters, and for any sudden changes in the political landscape (like a scandal) that might arrive before election day. Polls, on the other hand, are attempting to get a sense of the electorate itself — meaning that they tend to be snapshots of the present moment. Polls at the moment may seem to show that Democrats winning 49 seats is their likely electoral ceiling, but Polymarket clearly thinks that polls will continue to get bluer as the election approaches.
Additionally, one must consider who is betting on prediction markets. Though claiming to be more accurate than polls, they run into the same issue that most polls work to avoid: failing to offer a representative sample of the opinions of all Americans. More than 70 percent of Polymarket visitors are male, and the majority of visitors are below the age of 45. Clearly, Polymarket is an average of what these people are willing to bet money on, not how the American electorate in specific places plans to vote. Most of the time, the “common sense” of Polymarket feels out the correct choice, but as demonstrated in the Texas race, Polymarket can absolutely be wrong.
This is a problem. Not necessarily that Polymarket can be wrong, but why it’s wrong. Polymarket and other betting apps are already more influential than they might appear. High-profile political leaders are aware of the predictions made on Polymarket — White House Press Secretary Karoline Leavitt once ended a briefing almost exactly at the 64 minute mark, seemingly taking advantage of a Polymarket bet predicting with 98 percent odds that the briefing would run longer than 65 minutes. One can imagine that political officials could begin using Polymarket as one of many metrics to gauge public opinion on certain issues.
But they wouldn’t truly be gauging public opinion; they would be gauging the conglomerated opinions of men between the ages of 16 and 45, the people who actually use Polymarket. That’s a problem because that group is subject to certain biases, particularly along partisan lines. Last year, a Pew Research poll found that young men as a whole had shifted right by 5 percent since 2020, putting them at 53 percent Republican or Republican-leaning. And while it’s impossible to be sure, it seems likely that the percentage of Polymarket bettors who align with Republicans is even higher. Thus, men who lean right likely have the greatest say in determining the futures predicted by Polymarket — and while they might not be betting strictly along partisan lines, nothing can take away the outsized influence that they possess.
If prediction markets become a measure of public opinion for the White House, or a means for an administration to gain some kind of vindication or approval for their actions, it will be privileging a certain section of the population above others. It will give the section of the population that are bettors a greater say, whatever their political beliefs or demographic might look like. At the end of the day, it’s impossible for Polymarket or any other betting website to accurately reflect the whole feeling of the country.
So the trouble is that these prediction markets fail to represent the views of most Americans. But the trouble doesn’t end there. Prediction markets also absolutely put opportunities on the table for powerful people to manipulate events in a direction favorable to themselves. If Polymarket is leaning in a direction that turns out to be wrong, those with inside knowledge and financial resources can easily make a substantial profit by betting against that outcome. This doesn’t necessarily mean elections, of course — Polymarket allows its customers to bet on all manner of geopolitical developments, including whether there will be an American ceasefire with Iran by the end of the year. All that to say: online betting markets are currently ripe with opportunity for corrupt insider trading and manipulation by wealthy individuals. Without significant, immediate regulation of their operations, betting markets could quickly become the next iteration of the Trumpcoin crypto scam.
Last year, I advocated for a number of potential reforms aimed at ensuring that betting markets could function as relatively unbiased predictors of election outcomes (the most dramatic of which was nationalization). Having now seen just how wrong Polymarket can be, and having asked myself why that is, I’ve changed my mind. It is my belief that Polymarket cannot serve as a truly accurate metric for electoral victories, or for any other political development. In fact, it is these “other political developments” that scare me the most.
Polymarket recently removed a market betting on whether nuclear weapons would be detonated this year after internet backlash. Before the market was removed, bettors put the odds of a nuclear detonation around 24 percent. Just the existence of this market manufactures consent for a kind of humanitarian disaster that the world hasn’t seen in decades — and it isn’t alone. Markets rating the likelihood of a U.S. invasion of Cuba, a NATO-Russia military clash, or a ground offensive in Gaza launder all of these despicable actions as somehow acceptable, while simultaneously providing world leaders with a personal financial incentive to take these actions. Markets like these make Polymarket into just another cog in a machine legitimizing imperialism the world over.
If betting markets are to continue operating in the U.S., then the government has a moral duty to regulate them heavily, as they have proven that they are incapable of regulating themselves. Firstly, these apps should be required to limit each user to one account at a time, potentially by requiring proof of ID to create one. Secondly, government officials and their immediate families should be prohibited from placing bets on any markets where their position might give them undue advantage in determining the outcome, particularly all markets labelled “Political.” Third, a kind of oversight body along the lines of the Securities Exchange Commission dedicated to overseeing Polymarket trades should be formed at once, and instructed to prevent insider trading or market manipulation.
Additionally, betting market odds must not be allowed to become bywords for public opinion. News organizations that report odds, whether they be on the outcome of a Senate race or on what word Trump will say during his next address, should be required to indicate through a disclaimer that these odds do not provide an accurate representation of public opinion, and are not official polls. Lastly, and perhaps most importantly, certain types of markets must be entirely prohibited. Markets predicting the fall of a government, a military engagement, or the death of a political leader should be completely shut down, and bets on such outcomes should be prevented from being placed. While this might seem like an overabundance of caution, it is the best way to ensure that betting markets do not become mouthpieces for any kind of military campaign, whatever it may be.
Polymarket can be wrong, both in the factual sense and the moral sense. That latter problem shows no signs of solving itself; the only way to ensure that betting markets are ethical appears to be competent government oversight. While such a prospect seems unlikely under the present administration, Polymarket currently puts the odds of a Democrat winning the White House in 2028 at 60 percent. Let us hope that it is a Democrat who is willing to do what needs to be done — and let us hope that, this time, Polymarket is correct.
Featured Image Source: Better Markets
