By Shankar Vedantam
Monday, May 19, 2008
Hillary Rodham Clinton has half a dozen good reasons she thinks she is the best Democratic candidate for president. They are called Pennsylvania and Ohio, Arkansas and Nevada, New Jersey and New Mexico -- states she has won in the Democratic primary contest.
"The bottom line is this," Clinton said last week, after winning a landslide victory in the West Virginia primary. "The White House is won in the swing states, and I am winning the swing states."
Barack Obama has half a dozen reasons he thinks he is best suited to take the fight to John McCain, the presumptive Republican nominee: They are called Louisiana and Mississippi, South Carolina and Texas, Colorado and Kansas -- traditionally Republican states where the senator from Illinois won Democratic primaries -- evidence, his supporters believe, of his crossover appeal to independents and Republicans.
There is an obvious fallacy in extrapolating primary results to the general election. Voters in a primary are mostly from the candidates' own party, whereas voters in the general election include voters from the other party. Empirical evidence on whether primaries predict general-election performance is sparse, but there is one notable exception: A political scientist at Stony Brook University in New York has found an intriguing connection between performance in the New Hampshire primary and in the general election.
Helmut Norpoth said his model can correctly identify the winner of the popular vote in every election since 1952, with the sole exception of John F. Kennedy's narrow victory over Richard Nixon in 1960.
The model should interest Democratic superdelegates weighing the choice between Clinton and Obama, because it is based neither on opinion polls that ask voters about hypothetical matchups between McCain and the Democratic candidates nor on general trends such as the state of the economy and whether Americans think their country is headed in the right direction.
Norpoth's model is unusual for several reasons. How can the views of partisans in any state, especially one as small as New Hampshire, say very much about the views of the country as a whole? Second, while New Hampshire's first-in-the-nation primary has long been known to play a powerful role in deciding the eventual Democratic and Republican nominee, Norpoth isn't particularly interested in whether a candidate wins the primary. Many politicians -- Bill Clinton in 1992 -- have lost the New Hampshire primary but gone on to win the White House.
What Norpoth cares about is the proportion of the vote a candidate wins in New Hampshire, relative to his or her strongest competitor. This number is not just a measure of individual popularity but also a reflection of how united a party is going into an election year. Norpoth thinks New Hampshire's results are potent because the state has many independents, and candidates who do well there are likely to appeal to independents in the general election.
In this year's New Hampshire primaries, Hillary Clinton won 39.1 percent of the Democratic vote, while Obama won 36.5 percent. Together, they won 75.6 percent of the overall Democratic vote. On the GOP side, presumptive nominee John McCain won 37.1 percent and his nearest competitor, Mitt Romney, won 31.6 percent. Together they won only 68.7 percent of the overall GOP vote -- the lower combined number suggests the GOP was less united at the time.
For Obama, the magic number produced by the model is 48 (36.5 divided by 75.6), while Clinton scores 52 (39.1 divided by 75.6). McCain scores 54 (37.1 divided by 68.7).
Norpoth's model shows that over the past several decades, successful presidential candidates from the party incumbent in the White House -- in this year, the GOP -- need a score of at least 56. Successful candidates from the opposition party need to cross 47.
Both Obama and Clinton, in other words, are above the threshold for victory predicted by the model, while McCain falls below.
"I think either Clinton or Obama is likely to beat McCain," Norpoth said, as he explained how his model also takes into account cyclical trends. When both the Democratic and Republican nominees make it above the model's threshold -- as happened in 2004 -- or fall below the threshold -- as happened in 2000 -- Norpoth takes other factors into account to break the tie.
"Obama wins by a hair -- 50.1 versus 49.9," Norpoth predicted about this year's election. "With Clinton, she does a little better -- 50.5 percent against 49.5 percent."
James E. Campbell, a political scientist at the University of Buffalo and the author of "The American Campaign," praised Norpoth's insight that early party unity can be a useful predictor of general-election performance. But he argued that the 2008 election might reveal the potential flaw in the model: The unusual primary season this year has left Democrats much less united now than they were during the New Hampshire race, and Republicans much more united.
"There is a lot of animosity that has been created over these last few months, and both Obama and Clinton are drawing from distinctive segments of the party, so it is unclear if the hard feelings can be set aside at the voter level," he said. "There are going to be a lot of women upset if Obama is the candidate, and the same will be true of black voters if Obama is not the nominee. That is going to make healing the party's wounds difficult."