Are the polls overstating Trump’s support?

The New York real estate mogul and reality television star paces the field of Republican presidential candidates nationally, garnering an average of 35.1 in the Huffington Post polling tracker. Rounding out the top four candidates are retired pediatric neurosurgeon Ben Carson, with 13.9 percent and Sens. Marco Rubio of Florida and Ted Cruz of Texas, with 12.6 percent and 11.4 percent, respectively.

But use the site’s custom filter to remove surveys conducted via the Internet, Interactive Voice Response, a combination of IVR and online and through automated telephone interviews, and the data looks significantly different. Using only polls conducted via interviews with live telephone operators, Trump’s still commanding position in the national GOP average drops to 27.2 percent; Carson’s number rises to 20.2 percent.

“The pattern is definitely meaningful, but like the effect itself, we don’t really know what the lesson is yet,” Huffington Post senior polling editor Natalie M. Jackson told the Washington Examiner on Monday.

Trump’s competitors aren’t surprised by this disparity. They believe it’s legitimate; it’s among the reasons why they are more optimistic about their prospects than might be expected given that Trump has led in most state and national polls since July, except for a recent period when it appeared he had been dethroned by Carson. Senior advisers to rival GOP campaigns say the polling has been influenced by several unique factors, including faulty methodology.

The blanket coverage of Trump on cable television news stations has obscured the rest of an otherwise deep field of contenders. When asked in polls why they are supporting Trump, many voters have said that they haven’t heard anything about the other candidates. That has begun to change with in uptick in advertising by, and media attention on, other candidates. This change could have a major impact on the race in the days ahead.

Then there are methodological issues.

Auto-dial polls tend to miss cell phone users, resulting in polls that skew older and whiter. Online polls tend to be opt-in surveys, ruining the element of random selection of respondents that many pollsters believe is necessary for accuracy. And perhaps most importantly, say many, though not all, pollsters, live-dial polls tend to be the most reliable way of determining who a likely Republican primary voter is, versus someone who identifies as one but won’t participate.

“It’s really not that surprising,” a Republican pollster who is advising one of Trump’s competitors said, on condition of anonymity, when asked if it was odd to see the New Yorker’s numbers drop when only live-dial polls were taken into account.

Professional pollsters and polling analysts differ on the value and accuracy of expensive, traditional live-dial surveys vs. cheaper online and automated polls. RealClearPolitics, for instance, excludes Internet polls from its widely reported-on national average, although it includes auto-dial surveys. Trump tops Carson 28.7 percent to 19.7 percent in the RealClearPolitics average, similar to Huffington Post’s average of 27.2 percent to 20.2 percent if all but live-dial polls are filtered out.

Frank Luntz, a Republican pollster, said conducting national surveys to gauge support for GOP primary candidates is a waste of time and reveals very little about the state of the 2016 nominating contest. What matters are the candidates’ standing in Iowa, New Hampshire, South Carolina and Nevada, the first four states, in that order, to cast caucus and primary votes.

Luntz’s other problem with the much-heralded national primary polls are the sample sizes — many rely on too few voters to draw their conclusions. This year, several national GOP polls have used sample sizes of around 400 or less. That’s not large enough to produce a dependable prediction. Indeed, a candidate polling at 5 percent could really be at 10 percent; a contender at 3 percent could be tied with one at 5 percent.

“I’ll make it simple,” Luntz said. “Trump is leading just about everywhere. Carson is second just about everywhere. Cruz and Rubio are battling for third just about everywhere, and everyone else is in single digits.”

Rubio and Cruz, who have been on the move in state and national polls, sit at 12.6 percent and 11.4 percent, respectively, in the Huffington Post’s unfiltered average of national GOP surveys. Those numbers don’t change all that much when only live-dial polls are averaged; in that case, Rubio is still in third place in the race at 12 percent, and Cruz remains in fourth at 9.6 percent.

Filtering out nontraditional polls has a much bigger impact in Iowa — at least that’s the case for Carson and Rubio. The Iowa caucuses, set for Feb. 1, will mark the beginning of the 2016 nominating season. In the Hawkeye State, before removing auto-dial, IVR and online surveys, Trump tops the field at 26.4 percent, followed by Carson at 23.6, Rubio at 13.3 percent and Cruz at 11 percent.

Averaging only live-dial polls of hypothetical Iowa caucus matchups and Carson leads with 26.6 percent, followed by Trump at 22.2 percent, Rubio at 14.9 percent and Cruz at 10.5 percent. Cruz is in third place in the RealClearPolitics.com average of Iowa polls, garnering 18.3 percent; Rubio trails in fourth at 12.3 percent. As multiple pollsters pointed out to the Examiner, the survey averages are not exact science.

“In looking to compare various survey research contact techniques and the results they produce, like automated calls versus live caller versus on-line, introducing the additional element of defining likely voter screens makes the comparisons tougher as the likely voter methodologies for screens could be as different as the contact techniques in the results they produce,” Republican pollster David Winston said. “It makes it harder to make apple to apple comparisons.”