The first GOP debates for the 2016 election will be held Thursday, and polls decided not only who's on stage tonight, but who sits where.
But how accurate are these polls, and do they serve voters or politicians?
Fox News is hosting the first primetime debate. It includes what the network says are the top ten Republican candidates on average in five major polls at this point.
"This is a bad use of public polls," said Lee Miringoff of the Marist Poll.
By averaging five national polls, they say Scott Walker is favored by 9.5 percent of those sampled. Ted Cruz, 5.5 percent. Marco Rubio, 5.3 percent.
"To carry this out to decimal points — any pollster knows — that's just beyond reason," Miringoff said.
He assembles the Marist Poll, used by McClatchey newspapers.
When it comes to the Republican field, he said the fraction of a percentage point dividing many of the candidates is beyond the margin of error of any survey.
Miringoff said polls provide a valuable public service, but the consumer should be wary.
"There is no magic wand. No magic formula," he said. "I think people should understand who the sponsor is, so people know who is a real organization making a serious attempt to collect data; that they make their results transparent."
The website Real Clear Politics, Miringoff said, is one place where the public Can look under the hood of a poll. Inside are the actual questions pollsters ask, as well as the order in which they were asked. Both, he said, can influence the final results of the survey and are important to examine.
"Does this pass the smell test?" Miringoff said. "Was this in any way a biased question, or were both sides equally represented in the polls? So the person was not being led in one way or another."
NBC, CBS, ABC, Fox, and their partner newspapers all run highly-regarded polls that survey hundreds of respondents to get their numbers, Miringoff said. But to really know about a poll - whether it really says the message attached to it - peel back the percentage and look at the questions underneath.
