Election forecasters are confronting a decade‑long string of missteps, and the latest midterm cycle offers no relief. Since President Trump’s surprise victory over Hillary Clinton in 2016, pollsters have struggled to capture the true level of his popular support. The 2020 presidential race produced the worst collective error in four decades, and the 2024 midterms saw pollsters again underestimate Trump’s backing for a third consecutive election.
Recent polling embarrassments
August 2026 highlighted the problem in two high‑profile Democratic primaries – the Michigan Senate race and the Wisconsin gubernatorial contest. Both polls showed double‑digit leads for the leading candidates that vanished once votes were counted. The false optimism sparked headlines questioning whether the public can ever trust polls again.
Compounding the issue, a 21‑year‑old recent graduate created fabricated poll results for races in Wisconsin, Nevada and Los Angeles, posting them on a website called Median Strategies. One bogus poll claimed Los Angeles Mayor Karen Bass held a double‑digit advantage in her re‑election bid; the mayor’s campaign even shared the data before it was exposed as fraudulent.
Prediction markets rise as an alternative
While traditional surveys falter, prediction markets such as Kalshi and Polymarket have grown in prominence. These platforms let participants wager on election outcomes, offering real‑time probability estimates. Polymarket’s founder, Shayne Coplan, has called prediction markets “the most accurate thing we have as mankind right now,” and he has openly criticized conventional polling, saying, “Nobody takes polling seriously anymore.”
During the 2024 presidential election, both Kalshi and Polymarket projected a Trump victory, whereas aggregated poll averages gave Vice President Kamala Harris a slight edge. After the election, Coplan urged voters to “trust the markets, not the polls.” Kalshi echoed that sentiment, claiming its forecasts “decisively outperform polls and traditional media.”
Mixed results for markets
Prediction markets are not infallible. In Wisconsin’s recent Democratic gubernatorial primary, Kalshi and Polymarket assigned Francesca Hong a 95‑96 % chance of winning, yet she lost narrowly. The episode underscores that while markets can provide valuable signals, they remain subject to uncertainty.
Technical challenges facing pollsters
Beyond external competition, pollsters grapple with low response rates, forcing them to rely on statistical adjustments known as “weighting.” These adjustments align raw data with demographic factors such as education, race, gender, party registration and past voting behavior. Political scientist Josh Clinton of Vanderbilt University warns that even reasonable weighting choices can shift poll outcomes dramatically.
Another emerging concern is “silicon sampling,” where artificial intelligence agents simulate human responses. Companies like Gallup are exploring AI‑generated data to deepen insights, but traditional pollsters remain wary of the ethical and methodological implications.
What this means for voters
For citizens trying to gauge the political landscape, the takeaway is clear: treat poll numbers with caution and consider multiple sources, including reputable prediction markets, when forming expectations. The persistent underestimation of President Trump’s support suggests a robust base that defies conventional survey methods, a sign of enduring enthusiasm among his supporters.
As the 2026 midterms approach, pollsters will need to refine their methodologies and address the credibility gap highlighted by critics such as The Atlantic, which described the 2024 polling environment as “quietly still bad.” Whether they can regain public trust remains an open question.
Original reporting: KTBS 3 (Shreveport) — read the source article.
