In a few short weeks, Alan Abramowitz will predict whether Barack Obama or Mitt Romney will win the popular vote for president — and he's almost certain to be right.
Abramowitz, a political science professor at Emory University, has correctly called every popular-vote winner since he began forecasting elections in 1992. He has come within two percentage points, on average, of predicting the winner's share of the vote in those elections, too, including Al Gore's super-squeaky majority in 2000.
How can he accurately estimate the outcome of an election more than three months ahead of time — before the conventions, the debates, and the twists and turns of the fall campaign? Primarily because Abramowitz's forecasting model disregards the fall campaign altogether. His method assumes something that political operatives, journalists, and candidates rarely do: Presidential campaigns don't matter much in determining winners and losers.
Despite all the noise from the campaign trail, factors beyond either candidate's control largely determine the result, according to this school of thought. So much is already baked into a presidential contest that even the best managed and most effective campaign (or the most incompetent one) can't move the needle too far.
This idea has been around since at least the 1940s, and it has been so thoroughly studied that it has its own wonky name, the Minimal Effects Model. Simply stated, the MEM says presidential campaigns have a highly limited effect on how people vote. Because of partisan loyalties and other structural factors, millions of voters have made up their minds long before the most intense electioneering begins, leaving only a disengaged few for the candidates to persuade.
"When you're in the middle of a campaign, there's a tendency for people, especially the media, to overestimate the importance of certain events," Abramowitz says, such as high-profile gaffes, vice presidential selections, and controversial ads. But "those things have no measurable impact on voters' decisions. The media are interested in getting people's attention, but a lot of the stories you read or see are focusing on things that are trivial. The way campaigns play out is largely determined by fundamentals."
The "fundamentals" are broad measures of the electorate's satisfaction. Abramowitz plugs just three variables into his forecasting model — the president's approval rating at midyear, economic growth in the second quarter, and whether either party is seeking a third consecutive term. (He gives the incumbent party's candidate a bump if the answer to the last is no.)
Other forecasters, such as James E. Campbell of the State University of New York at Buffalo, have achieved results similar to Abramowitz's by considering a sequence of Gallup polls several months before Election Day. Yale economist Ray C. Fair has accurately modeled 21 of the 24 elections since 1916 with a method that considers economic growth, inflation, and incumbency.
These models discount the things that political junkies obsess over. Remember the brief flap in June about President Obama's comment that "the private sector is doing fine"? As Dartmouth's Brendan Nyhan pointed out, Obama's approval ratings increased in the three days afterward. Likewise, there's no evidence that a Romney adviser hurt his boss with his Etch A Sketch comment.
Even more startling, this analysis essentially ignores the candidates. "The assumption is that the major parties basically nominate reasonable candidates that are both well funded and reasonably well organized," Campbell says. "Therefore, they tend to cancel each other out."
By this reasoning, developments in the final months of the campaign rarely make much of a difference. Absent the start or conclusion of a war, a massive economic shock, or the mythical "October surprise," the vast majority of voters don't zig or zag in the run-up to Election Day. According to Campbell, the candidates leading in the Gallup poll in late September have won in 14 of the past 15 elections.
Consider, for example, the last months of the 2008 campaign. The financial meltdown that fall turned what had been expected to be a close race into a relatively easy one for Obama. Although some attributed Obama's victory to a brilliantly waged campaign, the economic news in September and October, coupled with President George W. Bush's deep unpopularity, would have been decisive for just about any Democrat four years ago, says John Sides, a political scientist at George Washington University.
Academic election-forecast models rest on the notion that campaigns don't have much effect because there aren't many voters left for the candidates to persuade, especially as the election approaches. Even in the worst of times, both major-party candidates can count on at least 40 percent of the vote; in the best of times, neither can expect more than 60 percent, Campbell says. Defections to another party are rare; some research suggests that voters change their party identification about as often as their religious affiliation, which is to say not very often.
People who describe themselves as "leaning" toward one candidate months ahead of the election overwhelmingly end up voting for that candidate. About 91 percent of Obama's "leaners" went for him in 2008, Campbell says.
This implies that the pool of "persuadable" voters is more kiddie-size than Olympic. And it becomes even shallower as the few holdouts start making up their minds. As of early May, the last time Gallup asked, only 7 percent of registered voters in 12 swing states said they were undecided.
The campaigns implicitly recognize this but rarely mention how few voters they're actually fighting for. This year, most of the campaign activity will be concentrated in about eight swing states and perhaps a few hundred thousand votes. It's not much of an exaggeration to say that the campaigns are training their massive firepower on some guy named Larry in Ohio and somebody's Aunt Betty in Tallahassee or Tampa.
For the news media, this should be sobering. News from the campaign trail is valuable and worthwhile; millions of people want and need to know how the next president is comporting himself. But as Abramowitz notes, the media err regularly by reading more into ephemeral campaign events than voters do.
Of course, our modern campaign industrial complex isn't shutting down anytime soon. Campaigns do have some role in influencing the outcome. "Minimal effects," after all, doesn't mean no effect whatsoever. Even true believers in the Minimal Effects Model recognize that campaigns and candidates are important, albeit at the margins. And those margins can be important, even decisive. Just ask Al Gore.
Campbell and other political scientists say one job of a campaign is to motivate partisans. Candidates "prime" voters by framing the issues in ways that will cause people to go to the polls. They can affect what voters know and whether they vote.
As for the forecast this fall, stay tuned. Abramowitz is a few weeks away from cranking up his prediction machine, so he's not ready to call the race. But he doesn't need a mathematical model to tell him what even the most casual observer can already see: "It's going to be a close election."