In a much-discussed column this week, Huffington Post political editor Thomas Edsall reviews an ongoing debate among political scientists over whether the presidential election is headed for a "blowout" for Barack Obama or a very close race with John McCain.
One side of the debate is anchored by an essay from political scientists Alan Abramowitz, Thomas Mann and Larry Sabato. "Virtually all evidence... points to a comfortable Obama/Democratic Party victory in November," they argue.
The rejoinder from James E. Campbell, another professor of political science, argues that "dissatisfaction with President Bush does not necessarily translate into antipathy to Senator McCain" and that "McCain has the record they [voters] are looking for, and Obama does not."
One assumption implicit in this debate is the notion that the outcome is not only foreordained but knowable, that pollsters have the power to discern some hidden "scientific" truth about the eventual outcome of the race. We see that assumption reflected in the arguments of partisans on both sides, who routinely cite polls that confirm their assumptions about the outcome and regularly condemn contradictory polls as biased and flawed.
Yet the fundamental limitation of horse-race vote preference results at this stage in the campaign is that they make poor predictors of the outcome in November. Those results tell us something about voter preferences "if the election were held today," but those preferences are always subject to change. In his column, Edsall kindly links to the electoral vote map now posted on the front page of my Web site, Pollster.com, showing "states with 284 electoral college votes -- 14 more than the 270 needed to win" leaning or strongly favoring Obama. The map, Edsall tells us, shows "that the presidential election is all but over."
Obama's current electoral vote majority in our analysis rests on the 20 electoral votes from the state of Ohio, where our current combined trend average gives Obama a 3.5-percentage-point lead. That margin is just barely enough for a light blue "lean" Obama status. It would take only a minor shift (or one good poll for McCain) to move Ohio back to the yellow "toss-up" status that most handicappers believe it deserves. As such, the race in Ohio and nationwide is far from over.
So what can we conclude from the polls of July?
First, there is no doubt that Obama has a small but statistically significant lead, especially among Americans who describe themselves as registered voters. We have logged data from 31 non-overlapping national poll samples in July, and all but two showed Obama leading McCain by at least 1 point. The probability of that pattern occurring by chance alone, as my colleague Brian Schaffner points out, is virtually zero.
Second, we have seen two surveys in July -- one earlier in the month by ABC News and the Washington Post, and one this week from USA Today and Gallup -- both suggesting that the race is closer among the voters who have typically been most likely to vote in past elections.
However, we ought to be cautious about that result, and not only because it rests on just two surveys. The "science" of selecting likely voters derives from finding the most accurate ways of using data collected in late October to predict votes on Election Day. It is anyone's guess, however, how "accurate" those techniques are when applied to data collected in July.
Third, we know that many voters have not made up their minds. The NBC/Wall Street Journal [PDF] poll released last week had one voter in five as either totally undecided (12 percent) or "just leaning toward voting for" their preferred candidate (10 percent). And, as reviewed here last week, many Americans -- perhaps as many as 30 percent -- have not yet formed strong opinions about either candidate.
Perhaps my background as a campaign pollster creates a bias, but I come to this debate with the firm belief that campaigns matter. The television coverage of convention speeches watched by tens of millions will matter. The fall debates, and their far larger audiences, will matter. The hundreds of millions of dollars spent by the campaigns and countless hours spent by volunteers to register, persuade and ultimately turn out voters will matter. Polls will help us track the impact of those activities.
Arguments about what we can predict from the polls of July are both entertaining and inevitable, but the reality is that our measurements are inherently fuzzy. Barack Obama has the upper hand, to be sure, but the race is far from over.