Democrats are heading into the 3 November US midterm elections with high expectations, encouraged by favourable opinion polls, growing public dissatisfaction with President Donald Trump and mounting concerns over the cost of living.
Polling averages suggest the party is well positioned to regain control of the House of Representatives, while several closely contested Senate races have raised hopes of a broader electoral breakthrough.
However, a history of polling errors in recent US elections has prompted questions about whether the surveys accurately reflect the political mood, particularly in states where Republican candidates remain competitive, reports BBC.
Recent surveys have shown Democrats leading in several traditionally Republican states.
A New York Times poll found the Senate race in South Carolina, which Trump won by 18 percentage points in 2024, effectively tied. CNN surveys have also placed Democratic candidates ahead in Texas, Kansas and Maine.
The findings suggest a potentially significant shift in the political landscape.
But previous elections have demonstrated that polling leads do not necessarily translate into electoral victories.
Trump outperformed polling expectations in his presidential victories in 2016 and 2024.
Even in 2020, when he narrowly lost to Democrat Joe Biden, most surveys had indicated a considerably more comfortable Democratic victory.
More recent contests have also exposed the difficulties facing pollsters.
In Michigan’s Democratic Senate primary earlier this year, surveys suggested Abdul El-Sayed was heading for a double-digit victory, but his eventual winning margin was just one percentage point.
In Wisconsin’s Democratic gubernatorial primary, a candidate who had enjoyed a polling advantage approaching 20 percentage points lost by a fraction of a percentage point.
Changing patterns of communication and declining participation in surveys have made political polling increasingly difficult.
Clifford Young, president of Ipsos Public Affairs, said pollsters were finding it harder to reach Americans as more people abandoned landline telephones, screened calls and declined to participate in surveys.
He also highlighted the challenge of reaching conservative voters, particularly those who distrust established institutions.
“Conservatives, especially those on the far right, are much more anti-establishment by nature,” Young said. “And they don’t participate in anything. They hate the system. That’s the challenge.”
If Republican voters are systematically underrepresented in surveys, polling models could overestimate Democratic support unless researchers adequately account for the imbalance.
Another difficulty is predicting who will actually vote.
Trump’s political appeal has repeatedly mobilised voters who might otherwise have stayed away from the ballot box, making turnout patterns harder to anticipate.
Even relatively small polling errors can prove decisive in closely contested Senate races, where candidates are separated by only a few percentage points.
Despite the uncertainty, general election polling has been more reliable in recent midterm contests than in presidential elections.
In 2022, surveys suggested Democrats could avoid the sweeping electoral losses normally associated with an unpopular incumbent president. The party ultimately lost control of the House but maintained its Senate majority.
Unlike presidential elections, midterms do not feature Trump directly on the ballot.
His ability to mobilise less frequent voters therefore presents a different challenge for pollsters when estimating turnout.
Young also cautioned against treating unexpected primary results as evidence that general election polling is fundamentally inaccurate.
Primary contests are historically more volatile, with voters often making decisions much closer to election day.
Although Trump’s low approval ratings and economic dissatisfaction provide Democrats with political opportunities, several factors could prevent the party from achieving the gains suggested by national polling.
Much of the territory being contested was comfortably won by Trump in the 2024 presidential election, giving Republicans an established electoral advantage in several key states.
Changes to congressional district boundaries have also reduced the number of competitive House seats, making it harder for either party to secure large gains.
Meanwhile, deep political polarisation has left fewer voters willing to switch between parties.
This increasingly entrenched electorate limits the impact of national political trends on individual congressional races.
Republicans have also enjoyed a substantial fundraising advantage, providing them with hundreds of millions of dollars to spend on campaign advertising, including attacks against Democratic candidates in the final weeks.
These factors could narrow the gap between the parties even if national polling continues to favour Democrats.
Polling averages, rather than individual surveys, generally provide a more reliable indication of electoral trends because they reduce the influence of unusual results from individual polls.
At present, even polling errors comparable to those recorded in recent presidential elections would still suggest Democrats are positioned for a strong performance on 3 November.
However, polling inaccuracies can work in either direction.
In state and special elections last year, Democratic candidates outperformed many surveys as their supporters turned out in unexpectedly large numbers.
Trump’s decision to hold campaign rallies in the final days in states he won comfortably in 2024 has also encouraged Democrats to believe Republican candidates may be facing greater difficulties than previously expected.
Nevertheless, late shifts in voter preferences remain possible.
Undecided voters could prove decisive if they move overwhelmingly towards one party in closely contested races.
Young said pollsters were improving their methods but acknowledged that predicting human behaviour remained inherently difficult.
“I think we’re doing a better job,” he said.
“But part of the problem we always have is not just with the pollsters. Part of it is that we’re dealing with humans. Humans are not easy.”
