You can now read 10 articles in a month for free on BostonGlobe.com. Read as much as you want anywhere and anytime for just 99¢.

ideas

Wait, campaigns don’t work?

They have a function, yes — but, political scientists are finding, it’s not usually to decide who wins the election.

Jim Young/Reuters

It was not Mitt Romney’s best moment.

“There are 47 percent of the people who will vote for the president no matter what,” the former Massachusetts governor said at a private, $50,000-a-plate fund-raising dinner in May, as captured by a hidden video recording released to Mother Jones magazine and published online last week.

Continue reading below

What he said next became a political flashpoint, as he went on to characterize the 47 percent of Americans who don’t pay federal income tax as both deadbeats and Obama voters. (In fact, a sizeable chunk of that group is likely to vote for Romney.)

Still, that first simple assertion—he had already lost nearly half the voters in America—was shocking in its own right. Since when does a candidate for national office admit that he has no shot at the support of nearly half the country?

To political scientists, though Romney may have misidentified which 47 percent won’t vote for him, what he said wasn’t far off. As they estimate, somewhere around 43 percent of voters on each side are unbudgeable partisans, immune to even the wiliest charms of the opposing candidates.

And lurking behind this division is a surprising point that is becoming clearer to political scientists with every election: When it comes to choosing a president, campaigns matter far less than we think they do. The better political scientists get at forecasting election results, the more it appears that all this campaigning, these endless months of ads and gaffes and debates—it all accomplishes very little.

Justin Sullivan/Getty Images

For many Americans, this seems to defy our most basic understanding of how politics works. Politics and campaigns are nearly synonymous, especially during election years. Politics is the art of persuasion, as the cliche goes, and in the loudest political system in the world this persuasion comes in the form of thunderous speeches, emotional drop-ins at downtrodden Rust Belt diners, and seemingly trillions of red, white, and blue balloons. It’s cloaked in history—the whistle-stop railway tour, the grandeur of debates—and it gobbles up shocking amounts of money, with nearly half a billion dollars having already been spent by the campaigns this year, according to the Center for Responsive Politics (to say nothing of the explosion in outside funding sparked by the Citizens United ruling).

But over the years, political scientists, led by election forecasters, have increasingly found that nearly the entire outcome of a presidential election is determined in advance, by forces that have nothing to do with the campaign or even who is running—things like rate of GDP growth, the presence or absence of an incumbent, and said incumbent’s approval rating.

“Almost all politicians and journalists and citizens way exaggerate the effect of the campaign,” said Michael Lewis-Beck, a political scientist and election forecasting expert at the University of Iowa. “Most voters don’t change their minds, period.”

This realization helps explain the recent trend toward increasingly sophisticated microtargeting in presidential campaigns—if only a tiny slice of the pie is unclaimed, it’s vital to find that slice before your equally hungry opponent does. It also suggests campaigns exist for reasons that have nothing to do with convincing undecided Americans—and that, as we all endure the daily ups and downs of the quest for the White House, it’s time to reconsider what the real purposes of this grand political theater might be.

***

It is impossible to literally measure the effect of a campaign—to do that would require re-running the same election with and without campaigns. But in the last few decades, important insights have emerged from the discipline of presidential election forecasting, in which experts have sought to build models that predict presidential vote share by plugging in so-called fundamental variables. For example, one of the more successful forecasters, Emory political scientist Alan Abramowitz, has made his past predictions based on just three inputs: the second-quarter change in real GDP in the election year, the incumbent president’s approval rating at the end of June, and whether or not one of the candidates is a sitting president running for reelection. (This election, he’s adding a fourth connected to political polarization.)

The simplicity of Abramowitz’s model is the rule, not the exception. He and the other researchers with the best-known published models—political scientists like James Campbell and Lewis-Beck—have proven able to predict outcomes of presidential races rather accurately using just a few key variables, almost all of them connected directly to the economy or the incumbent’s popularity.

On average, these models allow forecasters to predict election results within 2 percentage points of the eventual actual vote shares. This accuracy suggests that in a sense campaigns are generic: any candidate in Romney’s position would likely be sounding similar themes. But it also suggests something else. Given that you can predict the election without looking at the specifics of the campaign at all, that means that neither the quirks of the candidate nor the particulars of the campaign ultimately do much to move the needle.

In developing their prediction techniques, many forecasters initially left campaign variables out of the mix because they’re so hard to pin down, and because the final predictions of most forecasts are made well before the most intense campaigning begins. Now some are going back and trying to figure out how much of the (rather slight) inaccuracy of past predictions can be ascribed to the campaigns’ effect on voters—but they’re not finding much.

“It is very hard to find independent effects of the campaign once you put in these core variables,” said Lewis-Beck. In an article published in the journal Foresight, he and Richard Nadeau reported that campaign effects account for roughly half the average error in the most well-regarded models, or approximately 1.1 percentage points. “Even that’s really squeezing the turnip, you know,” he said.

One reason it’s so easy to predict an election based on preexisting factors is that the United States is much more crisply divided than polls often indicate. According to Pew, as of this year 33 percent of registered voters surveyed said they considered themselves “independent” (with 35 percent identifying as Democrats and 28 percent as Republicans). But it’s well known in political science that most of those “independents” have already chosen a side. “You don’t have to push them very far to get them to say, ‘Well, you know, I lean toward one party or the other,’” said Charles Stewart, a professor of political science at MIT.

When you factor out these “fake” independents, you’re left with just 10 to 15 percent of Americans or so who are true, persuadable independents. And since these invariably don’t break precisely down the middle, the actual effect of campaigns is even smaller.

So how many elections have actually been won on the campaign trail? Campbell, in a paper from 2001, estimated that about 1 in 4 postwar American elections have been close enough for campaign effects to be decisive. (He believes that campaigns decided the Truman-Dewey race in 1948 and the Nixon-Kennedy race in 1960, and may have decided Reagan-Carter and Bush-Gore.) Lewis-Beck and Nadeau, based on their own 1.1-point estimate of campaign effects (per candidate, meaning they think campaigns alone can account for a 2.2-point spread), put the fraction of elections that can be swung by campaigns at about a third.

All this suggests that in somewhere between 67 and 75 percent of presidential elections, campaigns don’t matter—the fundamentals so solidly benefit one candidate or the other that the end result is determined well before the first ballot is cast, whether the campaigns are being run by Karl Rove or The Situation. (It’s worth pointing out that these models predict popular vote share, which doesn’t always correspond to the Electoral College winner, as fans of Al Gore can attest. But there is, it’s safe to say, an extremely high correlation between the two.)

So what does this tell us about Obama versus Romney? If the forecasters are right, things may well be close enough this time for the campaigns to swing the outcome. In October, Campbell will be publishing a symposium on election forecasts in the journal PS: Political Science & Politics. He allowed the University of Virginia Center for Politics to publish a sneak preview of the predictions, and the average predicted vote share for Obama was 50.2, suggesting a breakdown well within the 2.2-percentage-point campaign-effect range predicted by Lewis-Beck. This is no guarantee that things will be that close, of course—and in fact recent polling suggests Obama may have a bit more of a lead than that. Still, this may be one of those relatively infrequent years in which independent campaign effects could make the difference.

On the other hand, the campaign effect could simply push things further in Obama’s direction. “In this case, I think it’s possible, though it’s a little too early to conclude this, that Obama may win by a larger margin than expected (at least in my model) because of Romney’s poor campaign,” said Abramowitz in an e-mail; his model predicts Obama will win, 50.6 to 49.5. “But the effect would still be small, altering the final margin by maybe 2-3 points.”

***

If campaigns so rarely hold the power to alter an election, why are they becoming a larger, rather than a smaller, part of American politics—bigger, more expensive, and more attention-
grabbing?

In part, campaigns have become something of an arms race. Lewis-Beck pointed out that the reason the forecasting models work without factoring in campaigns is because they assume—more or less correctly, it appears—that campaigns cancel each other out. If a politician skipped campaigning altogether, allowing his opponent to sprint ahead in the quest for public attention, “You would tank. You would lose big,” Lewis-Beck said. It’s one thing to have two noisy campaigns; it’s another to have one noisy campaign and only deafening silence on the other side.

There’s also the fact that an age of endless spending brings with it endless opportunities for employment at every level of a campaign (and, ahem, for those covering it). “[C]ertainly people who make their living selling their campaign management services are going to focus on” the importance of campaigns rather than the immovable fundamentals, said Stewart.

For now, then, with courts removing donation limits to outside political action committees, candidates opting out of public campaign financing that limits spending, and day-to-day coverage of the elections that increasingly functions as national entertainment, the trend toward bigger and more bombastic campaigns doesn’t look to be reversing anytime soon. And even if campaigns rarely settle elections, some experts are quick to defend them as having a useful role.

“Campaigns are for establishing the mandate for the post-election,” wrote Paul Gronke, a political scientist at Reed College. “Whether or not voters actually chose candidates based on campaign appeals or campaign events, it’s almost inevitable that the winner(s) claim a post-election mandate based on what they said in the campaign.” Campaigns also, Gronke pointed out, mobilize partisans to vote or get involved.

“People aren’t listening to the arguments and choosing sides” during campaigns, said Stewart. “What they’re doing is more subtle than that. They’re convincing themselves that this candidate is for people like me, or that candidate is for people like me. It’s sort of like sizing up a potential lifetime mate—you don’t make these decisions on the spur of the moment.”

Gronke’s and Stewart’s takes would suggest that watching campaigns as a horse race is the wrong idea. But listening to what the candidates actually say, and watching who they’re trying to connect with, does matter as a window into where America is likely to be heading. Just don’t believe the hype that every little up and down could have a lasting impact—Romney’s “47 percent” gaffe included.

“Every little glitch, people think, ‘Oh, this is gonna change it,’ ” said Lewis-Beck of the Romney recording. “Yesterday was Libya, today it’s this. But very few people are going to change their minds—even though maybe they should—but they won’t. They just won’t do it. I bet in two weeks from now you will not be able to find a trace of an effect from this thing.”

Jesse Singal is a frequent contributor to the Globe and a master in public affairs student at Princeton’s Woodrow Wilson School of Public and International Affairs. He can be reached at jesse.singal@gmail.com.
Loading comments...

You have reached the limit of 10 free articles in a month

Stay informed with unlimited access to Boston’s trusted news source.

  • High-quality journalism from the region’s largest newsroom
  • Convenient access across all of your devices
  • Today’s Headlines daily newsletter
  • Subscriber-only access to exclusive offers, events, contests, eBooks, and more
  • Less than 25¢ a week