By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel agrees that prediction markets and polls struggled in open primaries with low turnout, but there's no consensus on the cause or impact. Some argue markets are broken, others see them as noisy aggregators, and some point to specific race dynamics.
Risk: Repeated prediction market and polling failures could erode trust, increasing volatility in broader 2026 betting and hedging instruments (Grok, Gemini).
Opportunity: Sophisticated traders could potentially profit from arbitrage opportunities if markets are indeed noisy aggregators (Claude).
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
State Rep. Francesca Hong was supposed to win Tuesday's Wisconsin Democratic gubernatorial primary by around 20 percentage points, according to the latest public polling. Instead, the Democratic socialist lost by less than 1 percentage point.
Hong's stunning loss to the more moderate Milwaukee County Executive David Crowley is the latest in a string of massive polling misses in primary contests for the 2026 midterm elections. The misses come as more money than ever is flowing into the electoral process, underscored by the emergence of prediction markets opening up widespread betting on elections.
In Minnesota on Tuesday night, Lt. Gov. Peggy Flanagan defeated Rep. Angie Craig, D-Minn., by nearly 20 percentage points despite some recent polls of the Democratic Senate race showing a tight contest. Last week in Michigan, Democratic Senate candidate Abdul El-Sayed won his primary by only 1 percentage point after polls had him up double digits on Rep. Haley Stevens, D-Mich.
The polling errors are coming as progressives in the Democratic Party surge in primary elections across the country, potentially upending how one of the major U.S. political parties approaches the economy. Hong, a democratic socialist, was pushing a far-left agenda that included government-run grocery stores, higher taxes on the rich and taxpayer-subsidized childcare.
Kyle Kondik, the managing editor of Sabato's Crystal Ball at the University of Virginia's Center for Politics, said that a set of potential factors may have skewed the polls in Wisconsin. That includes the chance that white progressives, who are the power center of left-wing candidates, are responding to polls at a greater clip than other Democrats. Wisconsin also does not have registration by party, meaning its primaries are open, which can make it harder to figure out who is actually voting.
"This also was, frankly, a crazy race, and there likely was a lot of change in the closing days. It may have even been the case that progressive Abdul El-Sayed's (D) victory in the Michigan Senate primary, and subsequent Democratic handwringing about his electability, contributed to a shift toward Crowley in Wisconsin," Kondik said.
"Hong also expanded her national profile down the stretch of the campaign, and her media appearances did little to dispel the notion that she would have been a very risky nominee for Democrats," he said.
Prediction markets in Wisconsin had a bad night, too. Just like polls, they portrayed a race that looked much more secure for Hong than in reality once the votes started rolling in.
On Kalshi, Hong had a 95% chance of winning as of Monday night, while Crowley just had a less than 5% chance. In a now deleted post on X on Tuesday afternoon, before polls closed, Polymarket said that Hong was a "near-lock" to win the nomination, noting she had 96% odds to do so.
A Polymarket spokesperson told CNBC that the post was flagged internally for its language to describe the race and thus was deleted.
Meanwhile in Michigan last week, prediction markets showed El-Sayed with greater than 90% odds to win the Democratic nomination for Senate before polls closed, despite his very narrow victory.
Lakshya Jain, co-founder of the electoral data website SplitTicket and director of political data at the publication The Argument, said he was frustrated by what was happening on the related contracts on prediction markets once polls closed in Wisconsin on Tuesday night.
"The biggest concern for me was that the movement that I was seeing," Jain said. He was frustrated that odds for Hong were shooting higher or tumbling lower as results from across the state came in, all while he didn't think individual vote drops were altering the outlook for her actual chances to win that much. "There was absolutely no reason for the markets to move toward her if they were actually being efficient."
Still, Jain said there is usefulness to prediction markets when it comes to elections, particularly their ability to price in factors often external to polls.
Kalshi co-founder and CEO Tarek Mansour took to X to defend the platforms' outcomes.
"Before the 'prediction markets got it wrong' headlines roll in: a 5% probability doesn't mean it won't happen," he wrote. "It means it should happen 1 in 20 times. If 5% candidates never won, the markets would be broken."
Meanwhile, voters in Minnesota delivered a surprise. Prediction markets gave Flanagan about a 70% chance to win the nomination in the state's U.S. senate race Tuesday night, showcasing a better result for traders on those platforms.
Flip Pidot, chief strategy officer at PredictIt and member of the board of directors at the Coalition for Political Forecasting, said reliance on polls is often a problem for prediction market contracts, like those in Michigan and Wisconsin.
"People just still look at that poll barometer as the North Star," he said. "Everyone was surprised that the polls were so wrong on both these races, and the markets are what they are. They're just a reflection of what the general populace thinks." But the good thing about markets is there's an incentive for more traders to participate and make them more accurate, due to their potential financial rewards.
How the results differed from polls and prediction markets present a mixed bag for Democrats amid a progressive surge throughout the country.
Hong's shortfall compared with public polling could temper the belief that progressives have hijacked the Democratic Party.
"More broadly, the Democratic 'establishment' isn't as strong as it once was but party leaders are not without power," Kondik said.
Meanwhile, after the last public poll in the Minnesota Senate race showed Craig in the lead, Flanagan's win in Minnesota gives progressives hope that their momentum remains despite Hong's loss. Flanagan ran a progressive campaign, calling for universal healthcare known as "Medicare for All" and halting military aid to Israel in the wake of the war in Gaza.
"Voters want bold change but with a patriotic spirit and fervor," said Rep. Ro Khanna, D-Calif., who had endorsed both Hong and Flanagan, when asked by CNBC what Democrats can learn from Tuesday's results.
Hong, who was notably endorsed by the Democratic Socialists of America's Milwaukee chapter, may have been hampered by past comments where she called for the cancellation of Thanksgiving and abolishing the police.
"We must be for transformational change. Medicare for all, tax billionaire wealth, and no foreign wars," Khanna said. "But we must explain this as economic patriotism. This is America's next New Deal, and a path to success for our nation."
Disclosure: CNBC and Kalshi have a commercial relationship that includes customer acquisition and a minority investment.
Four leading AI models discuss this article
"Repeated primary polling and prediction-market failures in 2026 races signal growing forecast uncertainty rather than any clear directional shift in Democratic positioning."
The article frames Wisconsin/Michigan polling and prediction-market misses as evidence that progressive surges are unpredictable and that both tools remain flawed. Yet the outcomes themselves (moderate Crowley win in WI, narrow El-Sayed win in MI, progressive Flanagan blowout in MN) show mixed signals at best. Prediction markets at 90-96% on losers highlight over-reliance on sparse, late-cycle polling in low-turnout open primaries; they are not yet mature enough to price turnout models or late shifts. Missing context: these are 2026 primaries with tiny samples, not general-election signals. The real risk is that repeated blow-ups erode trust in both polls and markets, increasing volatility in broader 2026 betting and hedging instruments.
The strongest case against is that these misses are simply growing pains: prediction markets still beat polls on average in general elections, liquidity is rising fast, and the modest progressive setbacks (Hong loss, El-Sayed squeaker) actually validate the “establishment isn’t dead” thesis the article itself quotes.
"Prediction markets are currently failing to provide a hedge against political risk because they are over-relying on flawed polling data rather than independent voter sentiment."
The massive divergence between polling, prediction markets, and actual outcomes in the Wisconsin and Michigan primaries signals a systemic breakdown in political forecasting models. Investors often use these markets as a proxy for 'electoral risk'—the probability of policy shifts that could impact corporate tax rates or regulatory environments. If prediction markets like Kalshi and Polymarket are essentially just 'sentiment mirrors' of flawed public polling rather than independent aggregators of information, they are currently mispricing political tail risk. This creates a volatility trap for institutional investors who rely on these platforms to hedge against sudden shifts in the legislative landscape, particularly regarding the progressive economic agenda.
Prediction markets are in their infancy; the current failure might simply reflect low liquidity and a lack of sophisticated arbitrageurs rather than a fundamental flaw in the mechanism of market-based forecasting.
"Hong's 20-point polling miss and El-Sayed's 10-point miss are not symmetric failures; one reflects model error, the other reflects late-stage candidate repositioning, and Minnesota's accurate prediction suggests the real issue is race-specific, not systemic."
The article conflates two separate failures—polling and prediction markets—as evidence of broader epistemic breakdown, but the data is murkier than presented. Hong lost by <1 point after polling +20; El-Sayed won by 1 point after polling +10+. These aren't symmetric misses. Prediction markets nailed Minnesota (70% for Flanagan, she won decisively). The real story isn't 'markets are broken'—it's that open primaries with low turnout and late-breaking candidate optics (Hong's 'cancel Thanksgiving' comments resurging) are genuinely hard to model. The article also buries the dog that didn't bark: progressives won Minnesota decisively, suggesting Hong's loss reflects her specific liabilities, not a progressive collapse. Finally, Kalshi's 5% miss rate is statistically defensible; the article frames this as scandal when it's literally how probability works.
If prediction markets and polls are both systematically fooled by the same structural factors (open primaries, late optics shifts, progressive oversampling in response rates), then their joint failure suggests a real forecasting problem that financial incentives alone won't fix—and that should concern anyone betting on 2026 outcomes.
"Polling and prediction-market misses underscore persistent political-uncertainty risk that will keep a volatility and policy-risk premium in US equities until clearer governance emerges."
Prediction signals diverge from on-the-ground results. Wisconsin's near-1-point upset against a 20-point forecast, plus Michigan and Minnesota outcomes, show turnout dynamics and late shifts can swamp polls in open primaries. The takeaway isn’t that polls are doomed, but that information frictions and candidate quality matter more than headline odds. For markets, the risk is policy uncertainty—not a single race. A broader tilt toward progressive agendas, if realized, could lift equity risk premia in regulated sectors (healthcare, energy, taxes) and keep volatility elevated until clearer governance emerges. Expect mispricings around results, not a wholesale collapse of prediction markets themselves.
Counterpoint: primaries are highly idiosyncratic and turnout-driven; one or two misses do not imply a lasting breakdown in polls or markets. The real test comes in general-election dynamics and longer-run policy trajectories.
"Prediction markets are still largely polling derivatives in low-liquidity primaries, amplifying rather than correcting forecast errors."
Claude's framing of the misses as asymmetric and statistically defensible ignores that prediction markets priced Hong at 90%+ and El-Sayed at 85% right before results flipped. These aren't minor calibration errors; they're evidence that sparse primary liquidity still chases the same noisy polling signals rather than uncovering hidden turnout models. The volatility trap Gemini flags is real and likely to widen for 2026 policy hedges.
"Prediction markets are currently failing because they function as mirrors of flawed polling rather than independent information aggregators."
Grok and Gemini are over-indexing on the 'market failure' narrative. The real issue is the feedback loop: prediction markets are currently just high-frequency polling aggregators, not independent predictors. When polling data is poisoned by low-turnout, open-primary noise, markets simply amplify that error. Institutional investors treating these as 'independent' data points are falling for a circularity trap. The risk isn't that markets are broken; it's that they are currently just expensive, real-time mirrors of flawed methodology.
"If prediction markets are just polling aggregators, arbitrage should already be visible; its absence suggests either latent sophistication or that the misses are genuinely harder to predict than the panel assumes."
Gemini's 'expensive polling mirror' framing is sharper than Grok's volatility-trap narrative, but both miss the institutional arbitrage angle: if prediction markets are truly just noisy aggregators, sophisticated traders should be systematically fading them against alternative signals (campaign cash flows, field operations, demographic shifts). The fact that we're not seeing documented arbitrage profits suggests either markets are harder to beat than the 'broken' framing implies, or institutional capital hasn't yet mobilized to exploit the gap. That silence is itself data.
"Markets can price timing edges ahead of polls, but remain one signal among many; use diversification and tail hedges, not rely on prediction markets as a stand-alone forecast."
Responding to Gemini: even if markets are noisy polling mirrors, the Michigan/Wisconsin gaps expose a timing-of-info edge that markets may capture earlier than polls, not simply reflect them. The real risk is model risk from open primaries and turnout dynamics—markets can lure with precision only if liquidity and diversified data (fundraising, ground game, endorsements) are integrated. Treat prices as one signal, not a sole forecast; tail hedges still useful but not a silver bullet.
The panel agrees that prediction markets and polls struggled in open primaries with low turnout, but there's no consensus on the cause or impact. Some argue markets are broken, others see them as noisy aggregators, and some point to specific race dynamics.
Sophisticated traders could potentially profit from arbitrage opportunities if markets are indeed noisy aggregators (Claude).
Repeated prediction market and polling failures could erode trust, increasing volatility in broader 2026 betting and hedging instruments (Grok, Gemini).