Is This Really a 1 Point Race?

Do you believe that the evangelical share of the electorate will skyrocket from 23 percent in 2004 to 44 percent on Nov. 4? Or that John McCain will clobber Barack Obama by 52 points among voters under 25?

If so, I've got some polls to sell you. They're right over here... next to this bridge.

It's one of America's favorite political pastimes, especially common as October comes to a close--cherry-picking the surveys that help you (whomever you are) tell the tale you want to tell.

Over the past 24 hours, the Drudge Report, the Politico Playbook and Mark Halperin's The Page--three of the main arbiters of Washington CW--have all linked to the latest national polls from IBD ("close!") and AP-Gfk ("shock!"). Why? Because they show Obama leading by one point (45-44 and 44-43, respectively)--and none of the interested parties has any incentive to disagree. McCain wants us to show he's still in the hunt. Obama wants his forces to keep fighting. And the media desperately desires the drama and tension of a photo finish. Hence the lede on Liz Sidoti's AP write-up: "The presidential race [has] tightened after the final debate." And the new graphic that has magically materialized on FOX News: "TIGHT RACE. AP POLL SHOWS VIRTUAL TIE."

Unfortunately, the same polls also show McCain crushing Obama in the sub-25 demographic (IBD) and benefiting from a 19 percentage point boost in evangelical turnout (AP). Given that Obama typically leads by 20 to 40 points among young voters and that McCain isn't particularly popular with the born-again crowd, those of us in the reality-based community--that is, those of us who actually want to figure out who's winning and by how much--can probably agree that AP and IBD aren't the best polls to go by.

But that raises the question: which stats should reality-based observers rely on? Here are three tips:

1. The Law of Averages. Don't obsess over the constant influx of individual polls. A lot of them--like AP, IBD and, for that matter, the NEWSWEEK survey from June that showed Obama winning by 15 points--are probably outliers. Follow the averages instead. RealClear Politics only includes nonpartisan firms in its average; Pollster includes everyone. Right now, both sites show Obama ahead by an average of slightly more than 7 percentage points among likely voters--his largest lead of the cycle. That's pretty solid evidence that the race hasn't tightened. What's more, both sites show Obama hovering above 50 percent--meaning that he'd win a popular-vote majority if the election were held today.

Still, when consulting averages you should remember that they lag behind events by design. The RCP average, for example, includes polls that were in the field last Thursday. Averages are still our best reflection of on-the-ground reality. It's just that they reflect a reality that's a few days in the past.

2. A Likely Story. Given that only 60 percent of eligible adults vote in a typical election, pollsters can't simply survey a random sample of adults and then publish the results. They have to at least attempt to determine which of those respondents are going to vote. Hence screening for "likely voters."

In the past, public-opinion organizations like Gallup would ask respondents a series of questions--How much thought have you given to the upcoming election for president: quite a lot, or only a little? How often would you say you vote: always, nearly always, part of the time or seldom? Do you, yourself, plan to vote in the presidential election this November, or not?--and then weigh the results to determine who was likely to show up on Election Day.

The hitch with this model, however, is that it allows the pollster to "mak[e] a determination as to how the voter will behave." As Nate Silver writes, "a voter can tell you that he's registered, tell you that he's certain to vote, tell you that he's very engaged by the election, tell you that he knows where his polling place is, etc., and still be excluded from the model if he hasn't voted in the past." In previous elections, this wasn't much of an issue. But based on massive gains in Democratic registrations, enormous increases in turnout among "unlikely" voting blocs during the Democratic primaries and a nearly 20 point enthusiasm gap between Democrats and Republicans, there's a strong chance that Obama could "expand" the electorate by driving new voters to the polls on Nov. 4.

In an attempt to account for the electorate's potentially new complexion, an increasing number of pollsters are now publishing results based solely on respondents' stated voting intentions. Yesterday, Silver divided the eight current national polls that list separate results for likely and registered voters into two clusters: one (on the top) that relies on a "traditional" model and one (on the bottom) that relies on the "expanded" model. As you can see, the traditional filter tends to strip about 5 points from Obama's lead:

If you believe that millions of newly registered Obama supporters--young people, African-Americans, Latinos, etc.--won't show up on Election Day, stick to the traditional likely voter model. But if you suspect that at least some of them will, it's worth including the expanded version in your calculations.

3. The States of Play. With less than two weeks to go, McCain and Obama are largely ignoring the national numbers and focusing on key battleground states. You probably should, too. As the New Republic's Noam Scheiber notes, "having an active campaign in a state makes a big difference... McCain is husbanding his resources for the absolute minimum number of electoral votes he needs to win ... [so] there's no reason to think he couldn't lose the popular vote by 2-3 points but still win Virginia by 1."

To see how the states are shaking out, check RealClear Politics Electoral College page from time to time. Right now, RCP shows Obama leading by an average of more than 5 points in states worth 306 electoral votes; factor in states he's winning by average of less than 5 points--Florida, Nevada, North Carolina, Missouri--and his tally rises to 364.

The important number to watch, however, is how many electoral votes (EVs) Obama is collecting in states where he averages more than 50 percent support--i.e., states he'd win even if every single undecided voter breaks for McCain. As of today, the Illinois senator is topping 50 in all of the Kerry states (252 EV) plus Iowa (7), New Mexico (5), Colorado (9) and Virginia (13)--for a grand total of 286 EVs, or 16 more than he needs to win. What's more, there are signs that Ohio might be breaking his way as well. The three polls that were in the field this week--Big10 Battleground, CNN/Time and Quinnipiac--show Obama leading McCain 53-41, 50-46 and 52-38, respectively. Note that all of Obama's numbers start with a "5."

As with national polls, states averages lag behind events. So there's a chance that McCain could still catch up--or be catching up right now. That said, there's simply no evidence so far that "the presidential race has tightened." In fact, much the opposite. Like the rest of you political junkies, I'll be staying tuned to see whether something changes. But I won't let any single poll--however "close"--"shock" me into believing a storyline that's not supported by the stats.