Primary Polls and Prediction Markets Fail in Key Midterm Races

What Happened

On Tuesday, August 11, 2026, State Representative Francesca Hong, a democratic socialist, unexpectedly lost the Wisconsin Democratic gubernatorial primary to Milwaukee County Executive David Crowley by less than 1 percentage point. This outcome sharply contrasted with public polling, which had projected Hong to win by approximately 20 percentage points. Simultaneously, prediction markets like Kalshi and Polymarket had assigned Hong a 95-96% chance of victory as late as Monday night. In Michigan last week, Democratic Senate candidate Abdul El-Sayed secured his primary win by a mere 1 percentage point, despite polls indicating a double-digit lead over Representative Haley Stevens. Conversely, Minnesota's Democratic Senate race saw Lieutenant Governor Peggy Flanagan defeat Representative Angie Craig by nearly 20 percentage points, an outcome that contradicted some recent polls suggesting a tight contest, though prediction markets gave Flanagan about a 70% chance.

These results represent a series of significant forecasting errors in the 2026 midterm primary elections. The discrepancies have drawn scrutiny to the methodologies of both traditional polling and the burgeoning prediction market industry, which has seen increased financial investment. Polymarket notably deleted a post on X on Tuesday afternoon that declared Hong a "near-lock" to win, citing internal flagging for its descriptive language. The varied outcomes across these three states highlight a complex and inconsistent performance by electoral forecasting tools.

What the Evidence Establishes

The evidence establishes a clear pattern of significant divergence between pre-election forecasts and actual primary results in multiple Democratic contests. In Wisconsin, Francesca Hong's projected 20-point lead evaporated into a narrow defeat, while prediction markets, including Kalshi and Polymarket, assigned her over a 90% probability of winning. Kyle Kondik, managing editor of Sabato's Crystal Ball, suggested several factors for the Wisconsin polling errors, including a potential over-response rate from white progressives and the state's open primary system, which lacks party registration. Kondik also posited that Abdul El-Sayed's narrow victory in Michigan and subsequent Democratic concerns about his electability might have contributed to a late shift towards Crowley in Wisconsin.

In Michigan, El-Sayed's 1-point win defied polls showing him up by double digits, with prediction markets also giving him over 90% odds. Lakshya Jain, co-founder of SplitTicket, expressed frustration with the real-time movement of prediction market odds in Wisconsin, stating there was "absolutely no reason for the markets to move toward her if they were actually being efficient" as results came in. This suggests that even as votes were counted, market efficiency was compromised. The Minnesota outcome, where Peggy Flanagan won by nearly 20 points despite some tight polls, indicates that not all progressive candidates faced the same forecasting challenges, with prediction markets performing better in that specific race.

Where the Accounts Conflict

While the fact of the polling and prediction market misses is undisputed, the interpretation of these failures presents conflicting accounts. Tarek Mansour, co-founder and CEO of Kalshi, defended prediction markets by stating, "a 5% probability doesn't mean it won't happen. It means it should happen 1 in 20 times." He argued that if 5% candidates never won, the markets would be broken, implying that Hong's loss, despite high odds, falls within statistical possibility. This perspective suggests that the markets were not necessarily 'wrong' but rather reflected a low-probability event occurring.

Conversely, Lakshya Jain of SplitTicket criticized the real-time behavior of prediction markets, particularly in Wisconsin. Jain observed erratic odds movements for Hong as results trickled in, which he believed did not accurately reflect her actual chances of winning. He contended that such movements indicated inefficiency rather than a robust pricing mechanism. Flip Pidot, chief strategy officer at PredictIt, offered another perspective, attributing prediction market inaccuracies to an over-reliance on flawed polls. Pidot stated, "People just still look at that poll barometer as the North Star," suggesting that prediction markets often mirror public perception, which itself is heavily influenced by polling data, rather than independently pricing in external factors effectively.

Context and Stakes

The recent primary results and the associated forecasting failures carry significant implications for the Democratic Party and the broader electoral landscape. The unexpected defeat of Francesca Hong in Wisconsin, a prominent democratic socialist, could temper the perception of an unchecked progressive surge within the party, as noted by Kyle Kondik. This outcome suggests that while the Democratic 'establishment' may be weaker than in previous cycles, party leaders still retain influence. The internal debate within the Democratic Party regarding the electability of progressive candidates, particularly after El-Sayed's narrow win in Michigan, is likely to intensify.

The discrepancies also raise questions about the efficacy and reliability of both traditional polling methods and prediction markets, especially as more money flows into electoral forecasting. For campaigns, strategists, and voters, the inability of these tools to accurately predict outcomes introduces a higher degree of uncertainty, potentially impacting resource allocation and voter engagement. Representative Ro Khanna, who endorsed both Hong and Flanagan, emphasized the need for Democrats to advocate for "transformational change" like Medicare for All and taxing billionaire wealth, but to frame it as "economic patriotism" and "America's next New Deal," suggesting a strategic pivot for progressive messaging in light of these mixed results.

What to Watch Next

Observers should monitor the public statements and internal adjustments made by major polling organizations and prediction market platforms. Following these high-profile misses, there will likely be an increased focus on refining methodologies, particularly concerning primary elections where voter turnout and demographic responses can be less predictable. Specifically, watch for any announcements from polling firms regarding changes to their weighting or sampling techniques in states with open primaries or significant progressive voter bases. The response from prediction market operators like Kalshi and Polymarket regarding their public communication strategies and internal algorithms will also be critical, especially concerning the use of definitive language like "near-lock."

Furthermore, the Democratic Party's strategic response to these mixed primary results will be a key indicator. Pay attention to how the party's national committees and prominent figures allocate resources and endorsements in upcoming primary contests. Any shift towards explicitly supporting more moderate candidates or, conversely, a renewed push to unify behind progressive platforms, will signal the party's interpretation of these outcomes. The messaging employed by progressive candidates in future races, particularly how they balance calls for "transformational change" with broader electability concerns, will also be a significant development to track.

Bottom Line

Recent Democratic primary elections in Wisconsin, Michigan, and Minnesota exposed significant limitations in the accuracy of both public opinion polls and prediction markets. Francesca Hong's unexpected loss in Wisconsin, despite overwhelming forecasts for her victory, underscores the challenges in predicting primary outcomes, particularly in open primary states and amidst evolving political demographics. While prediction markets showed some success in Minnesota, their overall performance, especially in Wisconsin and Michigan, mirrored the polling failures, prompting questions about their efficiency and reliance on traditional data sources.

These forecasting discrepancies will likely prompt re-evaluations within the polling industry and among prediction market operators, potentially leading to methodological adjustments and more cautious public communication. For the Democratic Party, the results present a mixed bag for the progressive movement, suggesting that while progressive momentum exists, it is not uniformly guaranteed and may face electability concerns. The outcomes necessitate a strategic reassessment for both political parties and electoral forecasting entities as the 2026 midterm elections progress.