It's an election year, and a presidential one at that. And that means that we're going to see literally thousands of "horse race polls" asking Americans to name their candidates of choice for the more than 10,000 elected offices on the ballots this year.
Most of the attention will, of course, be on the presidential race. We're already deep into the primary/caucus polls, from the surveys in a single state (which make some sense) to the national primary surveys (which make substantially less).
But this is not another diatribe against the flood of horse race polls and the news media's preoccupation with them. As Jack Germond of the Baltimore Sun writes, any political journalist who fails to tell his audience who's ahead shouldn't be out on the campaign trail. And it isn't all that difficult to figure out which candidate is in the lead. It's the pol with the bigger poll numbers, right?
To quote Hertz advertisements, "Not exactly."
Many election contests aren't close races: One of the candidates may have a huge advantage in voters' minds over his or her opponents. A November Los Angeles Times poll reported that George W. Bush was leading Albert Gore by a 55 percent to 40 percent margin in a hypothetical general election matchup. So Bush really was ahead of Gore in this poll, carefully and professionally done in accordance with the best practices of the polling industry, right? Absolutely: Bush was ahead. (This doesn't mean Bush is going to win, but more on that later.)
And just as clearly, another poll might show Gore and challenger Bill Bradley knotted at 45-45. That's "even," "a dead heat," "as close as it gets" - well, you get the picture.
But what about another Los Angeles Times poll in New Hampshire: Gore, 43 percent; Bradley, 42 percent. Does Gore actually have a 1 percentage-point lead? Is Gore ahead of Bradley in this survey of registered Democrats in New Hampshire? Should the headline writer pen "Gore Leads Bradley"?
Not based on this poll. This survey, based on 249 interviews, has a sampling error margin of about 6 percentage points either way, on each number. Gore's margin over Bradley in this poll is only 1 percentage point, far less than the sampling error margin for the poll. In this case, the correct interpretation of the poll is that the two candidates are closely bunched in the voters' views - that is, it's a tight race and neither holds a lead. A little harder to put into a headline, perhaps, but far more accurate.
So that is my Commandment No. 1: When the margin between the candidates is less than the sampling error margin, you should not conclude that one candidate leads the other.
But what about this poll: Bush has 44 percent of the Republicans in New Hampshire, while Arizona Sen. John McCain has 36 percent. The sampling error margin is, again, 6 percentage points. So does this fall outside the realm of Commandment No. 1? After all, the margin is greater than the sampling error. But remember: the sampling error applies to both numbers - Bush's 44 percent, and McCain's 36. So the 8-percentage point gap is still less than twice the total sampling error.
Still, can't you conclude that Bush is ahead of McCain in this poll? Yes, Bush "is ahead," "has an advantage," "holds an edge." Now there is a chance that McCain might actually be ahead of Bush, even given this poll. That chance is small, but it is real.
And that's Commandment No. 2: If one candidate leads another by more than the sampling error margin but less than twice the sampling error margin, you can say that candidate leads slightly. But recognize that you might be wrong.
Next, Commandment No. 3: If the gap between the candidates is more than twice the sampling error margin, you can conclude that one candidate is definitely leading the other in the voters' minds - at least for now.
Perhaps the most important Commandment of all is No. 4: No matter how big the lead, no pre-election poll says the race is over until one candidate has won. Only the voters can make that declarative statement, by casting their ballots.