The Pull of Polls

By Thomas Lang

If the 2004 presidential campaign coverage were a sick child, it would certainly be suffering from an unprecedented case of the national poll virus.

From the start of April through May 7, 15 separate organizations (usually a combination of a polling firm and media outlet) released at least 23 polls handicapping the 2004 presidential election. That stands at an average of one poll every 38.6 hours. (This number does not include the Rasmussen tracking poll. That poll, conducted by machines, is refreshed every day.)

Those worst-stricken by the virus, including Fox News/Opinion Dynamics and Gallup Poll, conducted three surveys in that time span, offering its followers a new outlook on the race an average of every 13 days.

To put it lightly, this feeding frenzy hardly goes unnoticed by the most cerebral of Washington insiders. The beltway was in a tizzy in mid-to-end April after three outlets released polls defying the conventional wisdom that expected Bush to suffer after a barrage of negative media attention born out of the 9/11 commission and continuing turbulence in Iraq. Bush, despite all predictions, led John Kerry in trial heats.

In reaction, on Monday April 24, The New York Times editorial page commissioned two highly-regarded Washington journalists, Josh Marshall and Ryan Lizza, to rationalize the puzzling poll numbers for its readership.

Ryan Lizza of The New Republic accurately captured the fury: “Polls are to Washington what box scores are to cities that have a baseball team: they are scrutinized obsessively by partisans. This week the news has been especially startling.”

Lizza suggested that perhaps, a full six months away from the election, that, “Democrats should pause before they give up — and Republicans shouldn’t celebrate quite yet.”

The Times’ op-ed page was not alone in its exaggerated response to the poll numbers. As Eric Alterman pointed out in a April 29 column, a headline in the Kansas City Star read “Bad News Just Rolls off Bush” and declared Bush a “president with a sheen of Ronald Reagan Teflon.” Fox New’s Brit Hume went even further, proclaiming Bush “on the rebound.”

All this sturm und drang was in response to three separate polls released by ABC News/Washington Post, USA Today/CNN/Gallup, and Investor’s Business Daily, which at most had Bush leading Kerry by four percent with a three percent margin of error.

Almost no one stopped to consider this; national polls have their limits in measuring likely election results.

Dick Bennett, president of the polling firm American Research Group, contends that national polls can give a sense of where the country stands on a particular election or issue, and Dan Payne, a Boston-based media consultant with Democratic ties, still believes that national polling results function as a “jet stream” that can pick up the country’s reaction to international or national events, such as an overseas terrorist bombing.

However, as Payne demonstrated in an April 24 column for the Boston Globe, relying on a national poll to size up the election is like relying on an average national temperature to get dressed every morning. “US mean for April is 67 degrees. But Thursday it was 80 degrees in Boston, 54 in Chicago, 43 in Denver. National numbers don’t show local conditions in swing states.”

Bennett, shares Payne’s outlook stating that “in a 50-state race” national polls “could be right on and they could be meaningless.” And he concedes that polling partnerships continue to find national polls attractive because they require less time to conduct than a state-by-state count.

Simply put, national polls do not account for the way we elect presidents — via the Electoral College. (If the president were chosen by a direct popular vote, then national polls would be ideal. But, of course, that’s not how it works.) Before 2000, the last candidate to claim the office without a majority of the popular vote was Grover Cleveland in 1888. Thus, according to Bennett, there existed an “iffy” assumption that national polling could be used to measure the presidential election.

There are circumstances where a national poll can be a signpost for further analysis. Bennett contends “a certain lead [in the national poll] would tend to indicate an electoral college lead.” That number is estimated at ten percent, according to John Gorman, president of Opinion Dynamics Corporation, a polling firm that has partnered with Fox News for this election season.

Since Kerry wrapped up the Democratic nomination in early March, neither candidate has had a lead near the ten percent margin suggested by Gorham. So far, a large lead for either candidate has hovered at the five percent mark.

As Payne sees it, voters are tired of reading headlines declaring the race an almost dead heat. “How many times can we be told the country is split down the middle?”

Thus, Payne advocates a closer look at the battleground polls. In his April 24 column, Payne pointed to evidence that showed Kerry gaining on Bush in a pre-selected set of battleground states. “Proving national temperature theory, ABC/Washington Post poll showed [a] much tighter race in swing states than rest of country. Kerry leads Bush 46-44 percent in 17 battleground states. Bush leads 48-43 nationally with Mr. Consumer swiping 6 points.”

Even so there is no consensus that battleground polls are any more efficient or accurate than national polls.

Bennett told Campaign Desk that when you “break it down [into battlegrounds] it’s almost as meaningless as the national” because its still an average that does not account for certain gains and losses in a “winner take all system.” For instance, if Bush started to accumulate a large lead in Wisconsin, a battleground state, it would drive his overall battleground number up, even though Bush couldn’t use those extra votes to help him win another battleground state.

The murky definition of just what is a battleground state complicates battleground polling. So far this election season battleground states have been deemed to be those where the campaigns spend the advertising money. Mostly, that money has been spent in states where the race in 2000 was decided by less than five percent of the vote.

However, pollsters are becoming increasingly skeptical that this is a prudent way to approach the 2004 race. Two of those pollsters — Richard Morin, director of poling for the Washington Post and Gary Langer, director of polling for ABC News — made their case in a May7 Washington Post editorial: “[H]istory suggests another outcome: that this election’s real battleground states will be different from those of 2000. And what’s virtually certain is this: Covering the last election, like fighting the last war, vastly increases the chances we’ll miss what really matters in this one.”

Yet the polling continues because “no one wants to take the chance of missing something,” says Gorman.

Peer pressure feeds polling frenzy. As Payne told Campaign Desk, no “self-respecting” news organization “wants to be left out” and “forced to run with someone else’s poll.”

Dan Okrent, public editor of The New York Times broached this topic in a January 4, 2004 column. “I wondered why a single poll — The Times’s own, co-sponsored by CBS — was itself considered news ([when] at least one other released around the same time showed substantially different results.)”

Okrent continued, “… when any news organization touts its own polls while failing to note reputable polls conducted by others, I pat my pocket to make sure my wallet is still there. This isn’t news; this is awfully close to promotion.”

Regardless of the value of polls, Bennett says, “The bottom line is a newspaper puts its poll on the front page because it sells. It’s human nature for a voter to want to know where [he or she] stands compared to other people.”

While polls may serve this function, it stands to reason that both the casual observer and political junkie would be better served by a handicap that actually handicapped the race. And with an estimated $15 to $45 spent per interviewee per poll (polls question anywhere from 400-1400 people), according to Bennett, Campaign Desk is curious why more money can’t be spent breaking down polls like the Associated Press’s Ron Fournier did this past Friday.

Fournier, with the help of Will Lester, used the latest AP/Ipsos poll to project an electoral vote count for Bush and Kerry. The two candidates tied with 205 electoral votes. A candidate needs 270 votes to win the Presidency. (Fournier could not assign 138 votes from 12 states because of a lack of polling.)

The results of Fournier’s state-by-state race were anything but conclusive. But it does suggest a better way to present poll results in the future. That won’t happen overnight; it’s highly unlikely that we will see CNN anchors citing the latest state-by-state poll anytime soon. But, since polling has become an integral part of campaign coverage, and we’re here to critique campaign coverage, we feel compelled to ask:

Wouldn’t that approach make more sense?

Thomas Lang was a writer at CJR Daily.