Poll watch: Three keys to interpreting a very close race

As President Obama and his Republican challenger, Mitt Romney, prepare for tonight’s final presidential debate, polls nationally and in battleground states show the men in a dead heat.

Since the weekend, nine major polling organizations have released national surveys. One, Gallup, shows Romney holding a six-point lead among likely voters. Another, by Investors Business Daily and the TIPP polling organization, shows Obama up by four points. The others are arrayed between those two, with the latest NBC/Wall St. Journal poll showing the race tied at 47% apiece.

Similar results indicating an extremely close race have come from key states. In Ohio, to cite the state that is probably most critical to the outcome, a CBS/Quinnipiac University poll released Monday found Obama with a 50%-45% lead, down from a 10-point margin last month. Other polls done by automated polling rather than live interviews showed the race tighter.

Elsewhere, Romney appears to be narrowly ahead in the Southern battleground states of Florida and North Carolina. Obama appears to have a small lead in Nevada and Wisconsin. Colorado and Virginia appear to be true tossups. Iowa and New Hampshire have to be considered tossups as well since there have been fewer polls in those states than in some of the larger battlegrounds.


INTERACTIVE: Battleground states map

As we’ve said before, each polling firm uses a slightly different methodology, and since there is no single right way to do a poll, the best thing is to look at an average of several surveys and not focus too intently on any one. Currently, the averages show an extremely close race. The poll-tracking model maintained by Stanford University political science professor Simon Jackman for the Huffington Post, for example, shows the two virtually tied, with Obama at 47.2% and Romney at 47%.

For those who want to go beyond the averages and try to delve into the conflicting numbers, here are three keys:

Cellphones make a big difference. Some polling firms call cellphones and others don’t.


Calling cellphones adds a lot to the cost of a poll. Moreover, automated polls, which are much cheaper than live interviews, by law cannot call cellphones.

Not calling cellphones introduces a potential source of error since people who have no land line differ from the rest of the population. Cell-only voters tend to be younger, are more likely to be members of minority groups and are somewhat more Democratic.

Firms that don’t call cellphones try to weight their samples to account for those differences, but getting that right is tricky. Even firms that do call cellphones differ on what percentage of calls they make to mobile versus land lines, introducing another potential variance among polls.

In recent polls, Obama has tended to do better in surveys that do call cellphones. That’s particularly important when looking at state-level polls, since those are the ones most likely to be done on the cheap.
Republicans appear to have an edge on enthusiasm, but the size of the gap matters a lot.

That same Gallup poll which shows Romney ahead 51%-45% among likely voters shows a much closer Romney margin, 48%-47%, among all registered voters.

The numbers differ among polls, but virtually all show Obama doing several points better among registered voters as a whole than among those considered most likely to turn out.

That’s not surprising, and it reflects a reality — not all registered voters actually vote, and Republicans do better among actual voters because their supporters tend to be older and more certain to turn out.

This time around, that usual GOP edge has increased a bit, at least since the first presidential debate, by Republican voters’ enthusiasm about the race. That matters when pollsters try to figure out who is a “likely voter,” because the questions that some firms, including Gallup, use to decide who is likely to vote use enthusiasm as one factor.


The Obama campaign’s much-vaunted field operation is designed to erase as much of that GOP edge as possible by locating Democratic voters and making sure they cast ballots. If the turnout operation works as well as Obama aides hope, the actual voter population will be more Democratic than some of the polls of likely voters would indicate.

Republicans have their own voter-turnout operations too, of course. In 2008, the Democratic turnout operation outperformed the Republicans. This time around, the GOP hopes to neutralize that Democratic advantage.

PHOTOS: Memorable presidential debate moments

In some states, polling in Spanish as well as English could matter.

One of the most important factors in this year’s election will be the Latino vote — both how big it is and how heavily it goes for Obama.

But polling Latinos is complicated by language. Most polls interview respondents only in English or limit the number of Spanish-language interviews they conduct. As with calling cellphones, calling in Spanish increases costs. But interviewing only in English can produce a skewed sample. Spanish-speaking Latinos tend to be more Democratic in their voting than are English-speaking Latinos.

All other things being equal, a poll conducted only in English will probably show Romney doing better than one conducted in both languages and could overstate his vote. That may not matter significantly in a national survey but could matter in some states with large Spanish-speaking populations, particularly Florida and Nevada and perhaps Colorado.

The latest ImpreMedia/Latino Decisions poll, which is conducted in both languages, shows Obama leading Romney by 71% to 20% among Latino voters nationally.  


Finally, remember that margin of error is a real thing, not just a footnote. If a poll appears to show three points of movement in one direction or another, but the margin of error is +/-4 percentage points, the apparent shift could be just statistical noise. The odds of a shift being real go up if numerous polls done by different organizations all show the same result.

DEBATE QUIZ: Who said it?

Follow Politics Now on Twitter and Facebook

Twitter: @DavidLauter