With only a week to go before the election, pollsters are busier than ever. But some are wondering if the polls can be trusted at all this year or whether their spurious results might affect the outcome of the race. In a "Politics" thought experiment published in January, Daniel Engber imagined an election without polls. The article is reprinted below.
We learned two things from Tuesday's Democratic primary in New Hampshire: First, the Hillary Clinton campaign is very much alive; second, the pre-election polls were virtually useless. According to the numbers, Barack Obama was on pace to win by a margin of five to 13 percentage points; instead, he lost by three.
But opinion polling might be much worse than inaccurate. It's easy to imagine that the polls themselves affect the outcome of the elections they're supposed to predict. Voters may be inclined to jump on the bandwagon of a candidate who appears to be cruising to victory. Or they may stay home if they think their favorite is either out of the running or coasting to an easy win. Many believe that ubiquitous horse-race coverage pushes second-tier candidates out of the picture—and that Joseph Biden, Chris Dodd, Bill Richardson, and Ron Paul are all suffering at the hands of meddlesome pollsters.
Is it true that we'd be better off without polls? What would happen if pre-election polls were banned altogether? Let's conduct a thought experiment to find out.
Our counter-factual may seem prima facie ridiculous, since any attempt to ban media organizations from presenting opinion polls would run afoul of the First Amendment. But the idea doesn't seem so far-fetched when you consider similar policies that have been implemented in democracies around the world. According to a survey conducted in 1997, 39 of 78 countries had some kind of rule against pre-election polling. In the Philippines, the government briefly banned election surveys in the 15 days leading up to a national election. India's Election Commission banned the broadcast of opinion polls for two weeks in 1998. Starting in 1977, France had a two-week embargo on opinion poll results, which was later reduced to two days. Even our neighbors to the north imposed a three-day embargo in the mid-1990s. (The rule was later overturned by Canada's Supreme Court.)
For our purposes, let's assume a ban that applies to releasing poll results, as opposed to conducting polls. Using the abandoned Canadian law as a guide, we'll imagine a situation in which it's illegal to "broadcast, publish, or disseminate the results of an opinion survey respecting how electors will vote in an election." We'll also assume that the embargo lasts for the entire election season, instead of just the 48 or 72 hours leading up to Election Day.
First of all, the numbers would leak out anyway. We know that the Canadian embargo was violated by Internet activists who published survey results via U.S. servers. In our experiment, we'd expect to see major media outlets holding to the gag rule, while bloggers provided a steady diet of rumors and reports about the latest numbers. Savvy campaign strategists might try to feed false or self-serving information into the blogosphere, but true political junkies could consult mainstream media sources based overseas. (In short, only highly motivated voters with Internet access would keep abreast of the polling data.)
Without polling numbers to drive their narrative, political journalists would turn to other, less direct measures of candidate strength. At the outset of campaign season, the front-runners would be designated according to the size of their war chests, and the number of endorsements each had racked up. Focus groups, man-on-the-street interviews, and even voter brain scans would get more media play. Political futures markets might become a central element of mainstream campaign coverage, rather than the fringe oddity they are today.
Reporters might pay more attention to the turnout at campaign events, totting up average attendance figures and reporting them as rough guides to candidate popularity. As a result, campaign operatives would have an incentive to attract the biggest crowds they possibly could—and to hire extras to fill seats.
Of course, things might be very much the same without pre-election polling. Reporters don't always follow the numbers. Toward the end of 2006 and the beginning of 2007, for example, John McCain was repeatedly described as the Republican "front-runner," despite the fact that 53 of 57 national polls taken during that period showed him trailing Rudy Giuliani.
It's also possible that the worried-over "bandwagon" effect may not exist at all. What data we have from political scientists has been somewhat equivocal on the matter: Voters don't seem to be drawn inexorably toward the leader in yesterday's poll. If the bandwagon effect does exist, it might get canceled out by an equal and opposite inclination—the so-called "underdog" effect.
Pollsters concede that the effect of pre-election surveys is more clear-cut in the primaries, where voters are more inclined to vote strategically than they are in the general election. (It's no accident that the rise of public-opinion polling in the 20th century coincided with the movement toward primary elections.) By that token, a world without polling might help lesser-known candidates stick around. Meanwhile, the results of each individual primary election would become even more important than they already are. In our hypothetical scenario, the results in Iowa would be the first widely reported numbers of the entire race. The rankings established in Iowa would create a pecking order lasting all the way until New Hampshire—with no poll numbers in the meantime that might back up or deny the existence of any kind of "bounce."
In real life, pre-election polls seem to affect voter turnout in two ways. An apparent rout might make the outcome of an election seem like a foregone conclusion, leading voters to stay home. But polls showing a tight race tend to excite voters, and make them more likely to participate. We expect these effects to show up most acutely among young voters with a modest interest in politics—the kind who are interested enough to see the polls, but not fanatical about supporting their candidate.
If that's true, polling might have hurt Barack Obama's chances in New Hampshire. The polls leading up to Tuesday's Democratic primary suggested a decisive victory for the senator. But the polls may have depressed turnout among the young voters who were most likely to support him. (Independent voters might also have decided to vote for McCain, in what was thought to be the closer race.) So, in a world without pre-election polling, Obama would have had an easier time fending off a late surge from his opponents—but he'd have a harder time prevailing in close contests in the future.
The possibilities go on and on. But now it's your turn: What do you think would happen if our election cycle were spared the endless opinion polls? Would campaign coverage look any different? Or maybe the candidates themselves would change their messages, and their approach. …