Sociologists in the U.S. got their predictions for the congressional elections wrong, according to the FT
American sociologists warn that the results of the November 2026 U.S. congressional midterm elections may be difficult to predict. This is due to several recent primaries, the results of which differed significantly from earlier polls.
One of the most telling examples was the Democratic Party primary for the governorship of Wisconsin.
On Tuesday, August 11, David Crowley unexpectedly defeated Democratic socialist Francesca Hong, beating her by just 0.5%.
At the same time, opinion polls prior to the vote had painted a completely different picture: Crowley trailed Hong by approximately 22%.
This result was a positive sign for the party establishment, as a moderate candidate defeated a representative of the Democratic Party’s progressive wing. However, the huge gap between the predictions and the actual result called into question the accuracy of modern models.
A similar situation had previously occurred in the Democratic Party primary in Michigan.
Experts explain that primaries are more difficult to predict due to the relatively small number of voters who participate in them.
“Conducting polls during primaries is very difficult. When one candidate garners support among the group of people most likely to participate in polls—those with higher education, interested in politics, and motivated—it creates ideal conditions for error,” said Joshua Clinton of the Center for the Study of Democratic Institutions.
According to him, similar problems may arise during the midterm elections in November. American participation in polls is declining, partisan polarization is intensifying, and the turnout models used in 2024 may no longer work.
At the same time, Francesca Hong’s unexpected defeat cannot be explained solely by pollsters’ errors.
Toward the end of the campaign, an old social media post of hers went viral, in which she called for the abolition of Thanksgiving. The scandal broke too late, so its impact may not have been reflected in the poll results.
Hong was also criticized after stating that her son, who is of mixed heritage, supposedly gives her a certain “affinity with white people.”
In addition, the candidate may have lost some support due to the publicity surrounding radical statements made by other representatives of the democratic socialist movement.
Another factor was the shift in polling methods.
Due to low response rates and the high cost of telephone surveys, pollsters are increasingly turning to online surveys. This approach is cheaper, but its participants may differ significantly from the average voter.
People who voluntarily agree to participate in online surveys are usually more interested in politics.
The proliferation of large language models poses an additional problem, as bots can participate in online surveys.
Sociologists also have to estimate how many of those surveyed will actually turn out to vote, and then adjust the statistical weight of the responses based on these assumptions.
This is particularly difficult during primaries, where the traditional division of the electorate into Democrats and Republicans does not apply.
“In primaries, you lose your most important variable—party affiliation. That’s why conducting polls during primaries becomes really difficult,” noted Christopher Borick, a polling expert at Lehigh University.
Problems with predicting results have been evident in the past as well.
Last year, three weeks before the Democratic Party primary for New York City mayor, average poll results gave Andrew Cuomo a 10.5% lead over Zohran Mamdani.
In reality, however, Cuomo lost by nearly 13%.
One reason was the unprecedented turnout of young voters, which disrupted the models pollsters had relied on.
“What are the weighting factors for the New York election? You don’t know who’s going to show up to vote. They just make it up,” said Michael Bailey, a polling expert at Georgetown University.
Problems with the accuracy of polls in the U.S. have existed for over a decade.
Donald Trump’s victory in the 2016 presidential election demonstrated that Republican voters with lower levels of education—who make up a significant portion of his electorate—are less likely to agree to participate in polls.
After that, pollsters began to assign greater statistical weight to the responses of underrepresented groups. However, this did not fully resolve the problem.
“The Republicans who participate in polls are not Trump’s Republicans. So what we’re accounting for—or how we’re accounting for these people—doesn’t actually solve the problem,” noted Ashley Amaya of the Pew Research Center.
Despite all the shortcomings, experts are not calling for an end to opinion polls.
“Yes, polls have many problems, but if you abandon them, you’ll lose any chance of understanding what ordinary people think,” Clinton emphasized.
This was reported by the Financial Times.
Earlier, U.S. President Donald Trump stated that pro-Israel lobbying groups no longer wield the same influence in Congress and Washington’s political circles as they did about 20–25 years ago. He made these remarks in a pre-recorded interview for the program “America Speaks” on the Real America’s Voice network, according to Al Jazeera English.
Republican U.S. Congressman Andy Ogles lost the Tennessee primary to former State Commissioner of Agriculture Charlie Hatcher. Ogls suffered the defeat despite the support of U.S. President Donald Trump, Al Jazeera English reports.
