Five public polls published during the first week of October in the Kansas races for U.S. Senate and Governor.

Either something strange is happening in the Kansas electorate, or there’s something wrong with recent polling in the top two races in Kansas.

Based on polls released just in the past week, Republican Sen. Pat Roberts could be five points up, or 10 points down, or anywhere in between, in his race against independent Greg Orman

And Gov. Sam Brownback could be ahead of Democrat Paul Davis by as many as six points, or down by as many as five. Take your pick.

What’s especially striking about the polls is that they contradict what had been a pretty set pattern for most of the campaign, really going back to before the Aug. 5 primary: Roberts was trailing either Orman or Democrat Chad Taylor, or the combination of the two; and Brownback was running about five points behind Democrat Paul Davis.

So, either something has happened recently in the Kansas electorate causing a shift of some kind that is beyond measurement, or there’s something wrong with the new polls.

“Voters are fickle, but not that fickle,” said Patrick Miller, a polling expert at Kansas University’s Department of Political Science.

Bob Beatty, a political science professor at Washburn University in Topeka, agreed.

“We sort of have a little polling conflict,” he said. “They may converge in the end, but right now there’s a little discordance.”

Miller views the latest polls with a healthy dose of skepticism, especially those from the major news networks that have little history of ever polling in Kansas before.

“FOX/NBC/CNN have not polled in KS in the last decade and are likely less familiar with KS voter patterns, especially regarding GOP defections (e.g. Sebelius and Morrison got 30% of the GOP vote in 06),” Miller said in an email Thursday.

By contrast, he said, the other polls we’ve seen this year — from SurveyUSA, Public Policy Polling and Rasmussen Reports — have a longer track record of fairly accurate polling in Kansas.

Much of it has to do with methodology, Miller said. Most people are familiar with sampling methods – drawing names or phone numbers at random and separating people out either as “registered voters” or “likely voters.” But equally important is how they try to predict the population of people who will actually turn out to vote. That involves something called “weightings.”

“Weighting is complex, but it adjusts basically how much each individual respondent affects the poll outcome based on pollster beliefs about electorate composition and behavior,” he said. “I suspect that these national networks are looking at their U.S. polls and seeing how Republicans and Independents are voting on average for GOP candidates, and that’s affecting how they weight (Kansas) voters. So in effect, they are adjusting the numbers differently by network, but in a way that skews either the GOP or Ind vote from where the trend has been.”

Beatty said the weighting issue was especially evident in the CNN poll, which used a sample made up of 46 percent Republicans, 36 percent unaffiliated and 21 percent Democratic. That accurately reflects the registered voter population, Beatty said, but it’s not even close to the population that turns out on Election Day.

“Four years ago with Tom Holland (D-Baldwin City) at the top of the ticket, turnout was higher than that for Democrats,” Beatty said. “So this year CNN is saying fewer Democrats are going to vote? We’re not buying that.”

Miller noted that Public Policy Polling, a Democratic firm that has had a pro-GOP bias in its results, is expected to be polling again in Kansas this weekend.

“If PPP shows a move, I will begin to believe that voters are moving in a bipolar fashion overnight a la CNN/FOX,” Miller said. “If they don’t replicate CNN/FOX, then I will stick to my belief that this is a weighting issue.”