Earlier today, Gallup released a new set of poll results showing Mitt Romney leading President Obama by two points among likely voters, but another set of poll results showed the Republican challenger trailing Obama by three points. As Americans get bombarded with polling data in the closing weeks of the election, to which set of numbers should they be paying more attention?

“The closer we get to the election, the more you should focus on the likely voter model, because the entire purpose of the likely voter model is to isolate those most likely to vote and people are more aware of whether they are going to vote or not the closer they get,” Frank Newport, editor in chief of Gallup, told the Washington Examiner in an interview. “So I think it’s a good assessment now, and I think as we get closer it’ll be a better assessment.”

To determine whether somebody is likely to vote, Gallup asks respondents a series of seven questions – such as whether they are following the race closely; whether know where to vote; or whether they voted in the previous election – to determine the probability that they’ll vote again. The respondents who are considered “likely” are then filtered out to produce a separate set of results.

Newport said that Gallup made the decision long ago to begin screening out likely voters in October. In previous elections, the timing has varied. In 1992, for instance, he said Gallup didn’t make the switch until a week or two before the election, in other campaigns they made the switch as early as January.

“Early on (this time), we made the decision that we wanted to start reporting likely voters close to the election, but with enough time to get a feel for how they were trending,” he said. “And we made the decision we would start asking the questions on Oct. 1, which we did. Since we report seven-day averages we waited to accumulate enough interviews, and then we waited until Tuesday because that’s the day we’re going to be emphasizing our weekly assessments of the race between now and Election Day.”

Based on current polling, Newport described the race as “too close to call.”

This Thursday, Vice President Joe Biden will face off against Republican vice presidential nominee Paul Ryan in a highly anticipated debate. But Newport said Gallup has done an analysis of polling before and after every VP debate since 1976, which suggests the event has historically not had an effect on the race.  This was even true in highly publicized debates, such as Dan Quayle’s debates with Lloyd Bentsen and Al Gore in 1988 and 1992, and Biden’s debate with Sarah Palin in 2008.

“It is a fair conclusion that it’s hard to find evidence that a vice presidential debate changes the course of an election,” Newport said. A full report is due out on Wednesday.

Gallup conducts its polls by making calls between 5 p.m. and 9:30 p.m. every weekday, throughout the day Saturday (but not Saturday evening), and in the afternoon and evening on Sunday. Roughly half of the interviews are conducted on cell phone. The seven-day tracking poll is based on a simple average of seven days of polling results.

Since 1992, the Commission on Presidential Debates has hired Gallup to choose the undecided voters for the town hall debates.

For the purposes of picking the debate audience, Gallup pollsters call people who live near the debate site – in this case, it’s on Oct. 16 at Hofstra University in Hempstead, NY – and ask a series of questions that were pre-approved by both campaigns.

“If they qualify as an uncommitted voter, then we invite them to be a participant in the debate,” he explained.