The Washington Post and George Mason University began contacting voters last week in one of the country’s most competitive but least-polled Senate contests: the race in Kansas between Republican Sen. Roger Marshall and megachurch pastor Adam Hamilton, who is trying to become the first Democratic senator from the state since the Great Depression.
After more than 85,000 phone calls and nearly 50,000 text messages, a potential problem emerged.
Democrats were far more enthusiastic about taking the poll than Republicans were. Registered Democrats were more than twice as likely as registered Republicans to complete the survey after answering an interviewer’s call — 32 percent versus 12 percent — with a similar partisan gap when texting voters to complete the survey online.
This was a supersize version of a concern that has vexed pollsters for the last decade.
Partisan differences in answering polls are common, but not of this size. Similar polls of Michigan and Maine voters this fall found that cooperation rates among Democratic and Republican voters differed by less than four percentage points.
The Kansas poll was designed to minimize this disparity.
We drafted a shorter-than-usual questionnaire and conducted more than 8 in 10 interviews using live operators, who are more effective at contacting hard-to-reach voters.
Our survey firm, Braun Research, combed random samples from a statewide database to find response gaps among Republicans and made thousands of additional calls to compensate for them.
Then we weighted the poll results to match the Kansas electorate by party registration, 2024 turnout and vote choice, as well as a half-dozen factors including education, region and age.
But it’s possible those efforts fell short.
After aligning the sample with Kansas voters’ partisanship and demographics, respondents recalled supporting Donald Trump over Kamala Harris by seven percentage points in 2024, which is smaller than Trump’s 16-point winning margin in the state. Post-Schar School polls typically weight by 2024 vote choice. But even after doing so, registered independents in the poll reported voting for Harris over Trump by 25 percentage points, compared with a nearly even split in a 2024 AP VoteCast survey released after the election. The sample was weighted to reflect how partisans voted for president.
Given these patterns, there’s good reason to suspect that the poll overestimates Democratic candidates’ support. But The Post is publishing results in the interest of transparency and to highlight the challenges the poll faced.
In the U.S. Senate race, the Post-Schar School poll of 801 likely voters shows Hamilton leading Marshall 52 percent to 43 percent, with an additional 4 percent supporting Libertarian David Graham. However, the 3.9 percentage-point margin of error that applies to each candidate suggests a wide range of possibilities. Statistically, Marshall could be as high as 47 percent and Hamilton as low as 48 percent under traditional assumptions. The poll finds a narrower margin when asking which party voters want to control the U.S. Senate next year, with 51 percent favoring Democrats and 48 percent favoring Republicans.
But the nine-point Democratic advantage in the Senate race is a clear outlier from the other nine polls released since mid-September, which range from a five-point lead in polling sponsored by Hamilton’s campaign to a one-point edge for Marshall. A CNN-SSRS poll released Thursday found “no clear leader” with Hamilton at 49 percent and Marshall at 45 percent. All other polls show the candidates’ support within the range of sampling error.
Post reporter Teo Armus spent three days earlier this month in Salina, Kansas. Among more than two dozen voters he interviewed were MAGA loyalists supporting the incumbent, Democrats rallying behind Hamilton, and anti-Trump Republicans who were backing the Democrat, too.
Many of the Hamilton supporters insisted that some Republicans who voted for Trump two years ago would quietly cast their ballots for Hamilton or sit out the Senate election entirely, in part because of frustration over tariffs or rising diesel, fuel and fertilizer prices affecting the area, which depends on agriculture. But none of them could point to someone who was open about such a view.
“There is a heavy reluctance for people to speak openly about what they believe,” said Linda Salem, 62, an independent who hosted a fundraiser in Salina for Hamilton and said about a third of those at the event were registered Republicans.
But some change was noticeable.
“There was a time when everywhere you’d go you’d see pick-up trucks with the American flag proudly in the back and MAGA signs all over,” she said. “You don’t see that anymore.”
The Post-Schar School poll echoes warning signs for Republicans in other Kansas polling. A 58 percent majority of Kansas likely voters disapprove of Trump’s job performance, only slightly higher than the 53-57 percent disapproval in other recent polls.
But the new survey shows larger defections from the president’s base than other polls show, with 26 percent of likely voters who voted for him in 2024 now disapproving of his performance and a similar share of Trump voters supporting either Hamilton (18 percent) or Graham (5 percent) for Senate. Meanwhile, almost all Harris voters support Hamilton, as well as 71 percent of likely voters who didn’t vote in 2024.
Polls differ sharply on Marshall’s image, which was narrowly unfavorable in September polling but stands at 57 percent unfavorable to 32 percent favorable in the Post-Schar School survey. Most other polls have Marshall’s negative ratings outpacing positive ratings by single digits, though the CNN poll released Thursday found him underwater by 17 points.
Marshall has been subjected to heavy criticism since a New York Times report last month that he sued 700 patients for medical debts from 1998 to 2019, with 81 arrested for missing court dates. A spokeswoman for the senator has said that Marshall delivered more than 5,000 babies over his career and treated patients regardless of their ability to pay, including those on Medicaid.
Polling methods are changing to adapt to new challenges
The problems encountered in the Post-Schar School Kansas poll reflect growing challenges for election polls in recent years.
Survey response rates have plunged in recent decades, but low response rates themselves don’t cause polls to misfire. The problem comes when some groups respond more, such as Democrats or older people, and others, like Republicans or certain minority groups, respond less.
These gaps have always been present but have become more consequential since Trump galvanized support among a swath of less politically engaged voters, who typically respond to polls at lower rates. Trump has also attacked polls as fraudulent and designed to lower his public standing, which may also make his supporters less inclined to respond.
Polls underestimated Trump’s support by three to five percentage points in 2016, 2020 and 2024. At the same time, polls did not consistently overestimate Republican support in the 2018 and 2022 midterm elections.
After 2020, many polls, including The Post’s, adopted additional methods to guard against the threat of nonresponse bias, leveraging voter files with extensive information on past turnout and demographics to help ensure more representative samples. Polls also began weighting samples based on which candidate voters supported in the previous presidential election. Studies found this was effective in reducing errors in 2024, although polls still underestimated Trump’s support.
The methods used in the Post-Schar School poll proved fairly accurate in 2024, with an average error of 2.2 percentage points, while the partnership’s Virginia polls accurately showed Democrat Abigail Spanberger’s double-digit advantage in the 2025 gubernatorial election and the narrow passage of Democrats’ redistricting referendum this spring. This year’s midterm elections provide another important test, with post-election analysis pointing to ways they can be improved.
“The partisan enthusiasm gap likely explains the disconnect between Democratic and Republican willingness to respond,” said Mark Rozell, dean of the Schar School of Policy and Government at George Mason University, which co-sponsored the poll. “With an unpopular president, inflation, a war that Trump voters never asked for, the lack of GOP enthusiasm is not surprising. Are Republican voters this late in the election cycle feeling so down about their party’s chances that they just don’t want to talk about it?”
Teo Armus contributed to this report.
The poll was conducted by The Post and George Mason University’s Schar School of Policy and Government from Oct. 22-26 among a random sample of 801 likely voters in Kansas drawn from a statewide voter database. Live-caller interviewers reached respondents on cellphones (74 percent) and landlines (9 percent); 17 percent of respondents were reached via text message and invited to take the survey online. The overall margin of error is plus or minus 3.9 percentage points among likely voters.
The post We polled Kansas. Democrats wanted to talk. Republicans, not so much. appeared first on Washington Post.
