On Tuesday, November 14, exactly one week after Election Day (and with no President yet in sight), a notable though little-noted disclosure was made to the public. I do not mean the news that the federal judge in Florida had turned down the Republicans’ stop-the-hand-count motion, or the news that Bush’s lead in Florida was now 388 votes, or the news that a Florida state judge had waffled on Florida Secretary of State Katherine Harris’s decree that no county votes would be counted if reported after the 5 pm deadline that afternoon, or, for that matter, anything else that was happening in the murk of the Sunshine State. I mean the news that, according to a poll released by the Washington Post and ABC News, 45 percent of the public wanted George Bush to become President whereas only 44 percent wanted Al Gore to become President (6 percent wanted “neither,” 4 percent had no opinion and 1 percent wanted “other”). The claim was all the more striking in view of the hard contemporaneous fact that in the most recent count of the actual vote of November 7, Gore led Bush by a nationwide margin of 222,880 votes.
If anyone ever had doubts that politics in the United States is dominated by polling, this poll should put an end to them. A major poll was, in a manner of speaking, calling the election a full week after the vote–and reversing the known results.
The polls had been mercifully silent since the election. Many had good reason to be. Five of seven major ones had been “wrong” about the outcome of the election. That is, their final counts had failed to reflect the winner on Election Day (though some, it’s true, were within the margin of error). The New York Times/CBS “final” poll, which put Bush at 46 percent and Gore at 41 percent, had the margin wrong by more than five points and Gore’s final tally off by eight points. The Battleground poll, which gave Bush 50 percent to Gore’s 45 percent, likewise got the margin wrong by five points. Others were more modestly in error. CNN gave Bush 48 percent and Gore 46 percent; in the Washington Post it was Bush 48 and Gore 45; and in the Pew Research Center poll (with undecided voters counted), it was Bush 49, Gore 47. Only the Zogby poll, which put Gore ahead in the popular vote by 48 to 46 percent, and a CBS election-morning tracking poll, which gave Gore 45 percent and Bush 44 percent, picked the right winner in the popular vote, and with a margin close to the actual result. All in all, Gore’s victory in the popular vote came as a surprise. Of course, it’s not literally true that the polls were wrong, since there is a margin of error, and people can change their minds between the day of the poll and the election. On the other hand, election results are the only check on the accuracy of polling that there is–they are to polling what experimentation is to scientific hypothesis–and there is no reason to suppose that a poll whose final measure is 8 percentage points off the election result is not 8 percentage points off year in, year out.
Considering the decisive importance that polling had throughout the race in every aspect of the campaign, including media coverage, fundraising and campaign strategy (in the last few weeks of the election, hearts were lifting and falling on single-point fluctuations in poll numbers), these discrepancies deserved much reflection. The reason they did not get it was that on election night the magicians of public opinion went on to make even more egregious and momentous errors, by prematurely predicting the winner in Florida twice and the winner of the national election once. (The election-night calls made by the television networks, which in turn are based on exit polling done by a single, nearly anonymous firm, the Voter News Service, are not quite the same as opinion polling, since they record a deed–voting–rather than an opinion, but their use of sampling techniques to predict outcomes places them in the same general category as other polls.)