Democrats are leading in the latest polls for the US Senate elections, even in Republican strongholds such as Texas and Kansas, boosting their chances of regaining control of the chamber. But this lead raises a question prompted by previous election experiences: Do the figures reflect a genuine shift among voters, or are they once again overstating Democratic strength?
Surveys conducted by The New York Times and Siena University between October 23 and November 1 show Democratic candidates leading in Michigan, Texas, Kansas, Maine and Ohio, and holding a slight edge in Alaska and Iowa. To secure a majority, the party needs to retain its current seats and win four seats currently held by Republicans.
The Democratic momentum is linked to declining approval ratings for President Donald Trump and dissatisfaction with his handling of the economy and the war against Iran. Across the seven states included in the polls, his approval rating did not exceed 38 percent, despite having won five of them by wide margins in 2024.
Democrats also enjoy an advantage when it comes to voters' confidence in their ability to address the cost of living, while benefiting from the weakness of some Republican candidates. Analyzing the results, the newspaper's chief political analyst, Nate Cohn, argues that the Democrats' lead is broad enough to put them in a position to win the Senate, even if the polls underestimate Republican support by the same margin as they did in 2016 or 2020.
However, the possibility of polling errors cannot be ruled out.
The main problem is that Trump supporters are less likely to take part in opinion polls. In the latest surveys, registered white Democratic voters were about 30 percent more likely to respond than their Republican counterparts, a gap slightly wider than the one recorded ahead of the 2020 election.
This imbalance could result in an inaccurate representation of the electorate. In 2020, the final polls conducted by the two institutions underestimated support for Trump by five percentage points. Joe Biden won, but by a much narrower margin than the polls had predicted.
Still, differences in response rates do not necessarily mean that the results will be inaccurate. Polls in 2018 and 2022 were accurate despite higher Democratic participation. One possible explanation is that Trump supporters who are less likely to respond to polls may also be less likely to vote in midterm elections.
To address this imbalance, polling organizations use a technique known as "weighting," giving greater weight to responses from underrepresented groups. However, they may succeed in reaching the appropriate number of Trump supporters without accurately reflecting the diversity within that group: those who respond may be more inclined to vote Democratic than those who do not.
If too few Republicans participate in a poll, pollsters increase the influence of their responses on the final results to compensate for their underrepresentation. But this does not always solve the problem. Trump supporters who agree to respond may differ from those who refuse to participate. For example, they may be less committed to him and more willing to vote Democratic. In that case, simply increasing the weight of their responses is not enough to produce an accurate picture.
This summer, The Times/Siena adopted a new method for making these adjustments. It produced an average Democratic lead more than three percentage points smaller than the one the previous method would have shown.
But has the new approach made the results more accurate? That remains unclear. The new method may have corrected the overestimation of Democratic strength, or it may have underestimated their support too heavily. Republican support may also remain stronger than the polls suggest. The figures therefore give Democrats reasons for optimism, but only the ballot boxes will reveal how accurate they are.
