Why the Polls Drive Us Crazy (and Shouldn't)

There was a lot of talk last week about John McCain's "momentum" -- about the Republican brand rebounding.

And a dark cloud of gloom descended over many of those hoping to bring an end to the Bush era. "We've seen this before," was a common sentiment. A reader recently sent me a story, written during the lead-up to the 2004 election, about Kerry's seemingly indomitable lead in the polls -- a lead similar to Barack Obama's 8-point (average) advantage in the horse race a few weeks ago.

Then, this week, the storm clouds parted and the sun shone down on progressive America as Obama seemingly regained his mojo. Now he has surged back into the lead!

This emotional roller coaster is bad for one's psychic health and entirely unwarranted. The bigger picture is this: For about 10 days during the past 10 months -- after Sarah Palin's introduction to the country but before Americans got a good look at her beliefs -- McCain inched ahead of Obama in the national head-to-heads. Now, the tide appears to be turning back in Obama's favor: As the electorate has gotten enough of a look at Palin to distrust her, her once-high approval numbers have taken a nosedive. And McCain continues to say brilliant things like the economy is fundamentally sound and he won't meet with the dastardly prime minister of Spain.

More to the point, the significance of those head-to-head polls -- the yardstick featured in so much political reporting -- is completely overblown. The reality is that nobody knows what's going to happen in November. It's an unprecedented election in two ways: a black man is vying to become the president of the United States; and in the midst of an economic meltdown, voters are feeling an unprecedented degree of pessimism about the direction the country is heading.

Clearly, there is reason to believe that the "Bradley effect" -- the historical fact that some of those who tell pollsters that they're ready to vote for a person of color won't -- will have an impact on the election. How big will that impact be? Nobody knows.

Just as clearly, Obama has inspired a massive groundswell of interest, particularly among younger voters. Will they vote in numbers we've never seen before? Nobody knows (many have only a cell phone, and cell phone users are rarely polled).

At the same time, we know that in times of pronounced pessimism, people tend to vote the ruling party out of office. Yet, John McCain has spent a career carefully nurturing an image of independence from his party. Will he overcome the tarnished Republican brand? Nobody knows.

Those uncertainties are what ultimately make the polls almost meaningless. Pollsters don't offer us their raw data; they weight their results in various ways -- by age, party identification, past voting records, etc., in order to make their samples conform more or less with past experience. In an election for which we essentially have no past experience, it all becomes much more art than science.

Consider how they did during the primaries. Through the first five months of this year, Zogby's polling was off by an average of almost 7 points, Fox News by 4 points and Survey USA by 4.5 points. Mason-Dixon's average error was more than 8 points, and Rasmussen was off by more than 7.

Don't get me wrong -- those national polls, taken as a whole, are a good indication of broad trends in the mood of the electorate and are not without value. But the reality is that the nationwide polls we see being released almost every day are a poor indicator of what will happen come November.

Remember that we don't vote for president in the United States -- we vote state by state for electors who vote for the president. As I write, new polls show Obama up by 9 points in Michigan, 5 points in Pennsylvania and a couple of points in Ohio -- all crucial swing states.

The problem is that it's costly to poll each and every state, so many of those electoral college maps one sees in various publications are based on "trend estimates" -- on statisticians essentially reading the tea leaves, and often doing so using stale data. Of the 12 states that are in play this year according to The Politico, only three have had polls of registered voters within the past week, and all three of them show Obama in the lead. (There are more recent polls of "likely voters," but pollsters' "likely voter" models are a source of quite a bit of controversy and should be taken with more than just a grain of salt until just before the election.)

The second reason is that modern elections in America are largely mechanical affairs. It's true that Americans are not as deeply divided on the issues as they are commonly portrayed in the "red states versus blue states" narrative, but they are sharply divided in terms of partisan loyalty, or at least they have been during the past few elections. That means that winning will often come down to which campaign can better motivate its respective ground troops in a handful of swing districts in a handful of swing states. Obama's chief strategist, David Plouffe, told the Los Angeles Times that he ignores the national polls almost entirely because the campaign's focus is on registration and turnout in the 18 states they identified as the battleground (which is now down to maybe 10). "We stay laser-focused on these two factors," he told the Times. That may have been a bit of bravado, but there's no question that the Obama campaign has put an enormous emphasis on local organizing. As Daniel Nichanian wrote recently, "It increasingly looks as though this close election may be determined by investments on the ground, in local precincts across the country, an area of the campaign that Barack Obama's organization and supporters have excelled at." Nichanian added: "But this is not the type of edge pollsters pick up on."

There's quite a bit happening that polls have a hard time tracking, and much of it should be encouraging for Obama's supporters. The partisan identification of voters in many swing states (which I should point out is a less-than-rock-solid indication of voting, but important nonetheless) is looking very good for the Dems. Consider this graphic from Brian Schaffner at Pollster.com:


(click for larger version)

Not all states keep track of these data, but in six of the seven states included in the analysis, Democratic registration is on the rise, while the number of Republican loyalists appears to be declining.

But even those numbers don't tell the whole story. As Time Magazine reported, younger voters -- people under 35 -- a demographic in which Obama beats the socks off of McCain, are registering in pretty significant numbers. In 10 of the swing states, more than half of all new registrants this year are under 35.

Also not necessarily reflected in the partisan breakdowns are new citizens who will be eligible to vote for the first time. In 2006, there were 15 million naturalized immigrants who were eligible to vote. A huge effort on the part of immigrant-rights groups is expected to raise that number by 10 percent in 2008. If they can pull of the feat, that would be 1.5 million new voters, many of whom are clustered in swing states like New Mexico, Nevada, Florida and Colorado -- each of which Bush won by less than 5 points in 2004. Among Latino voters, 40 percent of whom voted for George Dubya in 2004, the key question is whether McCain can overcome the Republican brand, now seen by many as a party dominated by nativists. A recent poll of "Hispanic" preferences in four key battleground states showed that Obama is doing better among that group than Kerry did in 2004, and by a significant margin.

The other part of that equation is whether and whose votes are being counted. There's quite a bit of agreement across the political spectrum that America's electoral system is deeply flawed, and whether one believes its biggest problems are hackable voting machines and Republican voter suppression or Democrats registering people who are ineligible to vote, the polls aren't going to say whether it will be a fair contest.

The takeaway from all this is that those inclined to mope when the polls don't look good -- especially the national polls -- shouldn't. There's no point, because they mean little. If one wants to see an end of the Bush era, or a continuation of it, a better use of one's time is to get involved. Drive an old lady to her polling place. Put in some hours at a phone bank.

This is an American election, and the party that does a better job motivating its loyalists -- and perhaps expanding their numbers -- and protecting their votes will win. You don't need a poll to tell you that.