The polls on the upcoming election seem to suggest a very close contest between Democrats and Republicans for control of our nation. That is, if you go to any of the middle-of-the-road websites. The Conservative sites are spinning the polls to show that Republicans are certain to win big, whereas the Liberal sites have it surely going the other way. At times, reading these sites, I get the feeling that the folks doing the posting are just penning the fantasies that make them feel good, reality be damned.
The polls being quoted are usually John Zogby's polls. Perhaps because Mr. Zogby does not remain faceless and occasionally adds little nuggets of analysis, he tends to polarize more than most pollsters. On the partisan websites, the Zogby polls that show their guy with a big lead are unquestionably accurate, whereas the Zogby polls that show the other side's guy in the lead are flawed and perhaps deliberate, premeditated lies.
To cut through it all, it may be helpful to remember Zogby's predictions about the New York Senate race in the last election. The Sunday before voting day, one article said…
The Colgate/Zogby International Poll also paid carefully attention to the equally close senatorial race between Lazio and Clinton. "I think we're looking at a one point race," Zogby said. "I have no idea what's going to happen on Tuesday." Zogby observed that even although Lazio leads in the polls, it is not the lead that he would like to have over Clinton.
Two days later, Ms. Clinton defeated Mr. Lazio 56%-44%…hardly a "one point race."
I have, as always, a theory. I think you could go through the list of all the current elections and pick a winner in each race with no polling data whatsoever. In races where there's an incumbent, you write that he or she is ahead of his or her opponent, 47% to 41%. In races where there's no incumbent, you flip a coin between the Democrat and the Republican and put the winner of that coin-flip as ahead, 45% to 42%. If someone has a real big scandal getting attention, subtract four points. Then you say your poll has a margin of error of plus-or-minus three points…
…and, come Election Day, you'd be as accurate as any poll that actually goes out and talks to likely voters.