For all the signs of a dead heat in the horserace polls and strong hints of a victory for Senator John Kerry in exit polls on Election Day itself, the reelection of George W. Bush proved predictable from a model with a leading indicator available as far back as January. The forecast published in the October 2004 issue of PS was initially posted on January 29, 2004, the day after the New Hampshire Primary. It claimed, with 95% certainty, that President George W. Bush would defeat Democrat John Kerry in November. The forecast, which was not subject to any revision throughout the remainder of the election year, gave Bush 54.7% and Kerry 45.3% of the major-party vote. It missed the actual division of the major-party vote by a little more than one unit of the forecast error (2.7). All in all, a satisfactory performance of the model.

Before the election, the most common, near-unanimous, reaction of anyone upon hearing the forecast was, “Oh, I hope you are wrong.” Wishful thinking aside, a frequent retort was, “I'm sure you are wrong,” and “what do you bet?” Some of these skeptics, I admit, knew something about elections and pointed to undeniable warning signals. The most obvious was Bush's anemic approval rating, dipping into the low 40s in some polls and rarely settling above the 50-point mark. Only Truman in 1948 had managed to overcome such an obstacle. Another sign of danger was that a majority in polls believed the country was headed in the wrong direction. Plus, it was undeniable that the mood was beginning to sour over the war in Iraq, fed by an incessant stream of bad news. In midyear, Bush was trailing Kerry in the horserace polls. So how could this January forecast of a Bush victory possibly come true? Was it just dumb luck?

To recapitulate, the key predictor in my electoral model is performance in presidential primaries—by both major-party candidates, but only in New Hampshire for elections since 1952. In the Republican Party, Bush won the primary battle without a significant challenge. Any time a candidate of the party that controls the White House has done so, the odds are overwhelmingly in favor of that candidate's victory in the November election. This goes back as far as 1912, when presidential primaries were first used in large numbers. While Kerry won the primary battle in the Democratic Party, the model estimates assign the primary performance of the incumbent-party candidate far more weight in the November election than is true for the candidate of the out-party. The ratio is about 3 to 1. Voters are known to base their ultimate decisions far more on what they think about incumbents than challengers. About the only way the model could have forecast a Kerry victory in November would have been for the current president to have fared poorly in the primary battle—a repeat of his father's showing in 1992.

The overwhelming support George W. Bush garnered in the Republican Primary in New Hampshire was an uncanny bellwether of his party's unwavering support throughout the election year. Poll after poll of the general public showed self-identified Republicans planning to vote for Bush by better than 10-1 margins. On Election Day, according to the National Election Pool exit poll, a staggering 93% of Republicans cast their vote for Bush. That is a big reason why the forecast model got the winner right in 2004.

At the same time, these data also make clear that George W. Bush struggled with support from voters outside his party. In the end, he managed to attract 11% of Democrats and split the Independents about evenly with Kerry. On top of his near-unanimous backing from Republicans, that was enough to make the forecast of a Bush victory come true, but it certainly cut the predicted lead. My estimate was that for Bush to reach the forecast of 54.7% he needed to secure about six of 10 Independents. He fell short of this target. Whatever the explanation, it appears to be more a unique element of this election than a systematic factor that requires model revision. Note that in past elections incumbents with strong primary showings—such as Clinton in 1996, Reagan in 1984, Nixon in 1972, etc.—all succeeded in making deep inroads among the other partisans and Independents in November. Be that as it may, the model got it right in picking the winner of the 2004 election.