The poll hit like an earthquake in the run-up to the 2024 general election.
In Kansas — dependably Republican Kansas, of all places — a survey from Fort Hays State University showed Donald Trump only five points ahead of Kamala Harris in the presidential race. When combined with polls from Iowa and Ohio showing a surprisingly tight contest, the stage was seemingly set for Democratic overperformance.
That didn’t happen.
It so didn’t happen, in fact, that Trump won by more than 16 percentage points. Republicans in Kansas expanded their supermajorities in the Statehouse. Conservatives will control the State Board of Education. And those of us who followed the poll were left wondering what happened.
Thankfully, we now know. Brett Zollinger, director of the Docking Institute of Public Affairs at Fort Hays, wrote up a “2024 Kansas Speaks survey report addendum” that explains what went wrong and why. In short, the survey made one arguable mistake and likely missed shifts in public opinion because of the calendar. Those who dream of the day that Kansas becomes a blue state — or at least acquires a purplish tint — will benefit from paying close attention.
Zollinger highlights three areas that led to the polling miss.
First, the survey didn’t weigh by party identification. This might sound technical, but please bear with me. Those working in public opinion research traditionally balance their surveys in various ways. Imagine that you receive 100 responses to a poll, but only 20 of those responses come from college graduates. In Kansas, however, nearly 36% of adults have a bachelor’s degree or higher. The pollster will therefore give those 20 people a heft in calculating the responses, to more accurately reflect the state’s population as a whole.
Polls can be weighted in all sorts of ways. Age, gender, education, you name it. But the Fort Hays survey wasn’t weighted in one especially important way.
Zollinger elaborates: “Unlike in 2020, the 2024 Kansas Speaks survey data after weighting by age, gender, and education had an under-representation of Republicans (about 4%) and an over-representation of Democrats (about 7%). … Weighting by party ID was needed in 2024, particularly since Republicans and Democrats were so extremely polarized in their presidential candidate preference. After weighting by the best proxy for party ID available from the survey, Trump’s advantage among survey respondents increases to +10.5% among registered voters planning to vote and to +14.6% among all respondents planning to vote. The lack of a party ID question and then weighting by party ID is the largest factor accounting for the differential between the Kansas presidential vote and the candidate preference findings of the 2024 Kansas Speaks survey.”
In other words, the survey didn’t hear from enough Republicans and didn’t weight the GOP responses that it did have. To be fair, understanding what qualities to weigh in polling can be a difficult task. As recently as 2006, major public pollsters specifically opposed weighting based on partisan identification. But times change, and so do preferred methods of weighing polls.
In the 2016 presidential election, for instance, polls tended not to weigh by education level. If they had, Trump’s come-from-behind win would have been more apparent. Most public pollsters have started doing so in the years since.
The second problem Zollinger identifies is that “late deciders broke heavily toward Trump. 2024 Kansas Speaks surveying ended on October 16, almost three weeks prior to election day.”
That brings up an inescapable but neglected fact about all measures of public opinion: They exist in time. They tell what people think on the day those people are surveyed. That can change — and change dramatically — depending on events. While I personally found it difficult to imagine that late-deciding voters would trend Republican, that apparently happened in Kansas. Perhaps we shouldn’t be surprised, given the state’s durable partisan lean.
Finally, Zollinger notes: “In the last month before the 2024 election, October-over-September Republican voter registrations in Kansas outpaced Democrat voter registrations by greater than a 2 to 1 margin, with 15,710 Republican registrations in October compared to 6,336 Democrat registrations.”
In other words, Republicans in Kansas turned up to register in droves. Those new voters helped power Trump to his decisive win in the state.
I don’t want to dwell too much on the past. We have the election results, after all, and I’ve shared my takes. However, I think it’s worth taking a moment to ponder what this extra information from Fort Hays suggests about the path forward for progressive (and moderate, and sensible conservative) activists in Kansas.
Merely making sure that new voters register won’t be enough to change Kansas’ political lean. From the information cited above, a bunch of new registrations broke for Trump. Democrats have long believed that new voters and healthy voter turnout will magically benefit their side. I wonder if that was ever true, and it certainly doesn’t appear so now. New voters liked what they heard from Republicans.
Advocates also need to be realistic about dreams of turning Kansas blue. The state has a broad pragmatic streak and strong majorities favor abortion rights, expanded Medicaid and legalizing marijuana. But that doesn’t mean those majorities automatically vote for Democrats or a plethora of progressive priorities.
Joe Biden received 42% of the vote here in 2020. Kamala Harris received 41% in 2024. That bloc of Kansans appears stable, able to elect the stray Democratic U.S. House member or state lawmaker. In a gubernatorial race, it can even elect a governor if enough moderate Republicans join the fun. But doing even better will require rethinking assumptions, along with substantial investments of time and money.
Kansas remains a red state. Those supporting positive change and a different direction remain, too. How activists square those two facts will define our state in the years to come.
Excerpts or more from this article, originally published on Kansas Reflector appear in this post. Republished, with permission, under a Creative Commons License.
