In a presidential race as close as it is closely monitored, it's difficult to conceive of how candidates Barack Obama and John McCain could campaign without nationwide polling—or indeed how political pundits could analyze the race.
Used to ensure that messages are on target, point to potential problems, and decide where to ramp up efforts, polling can shape the contours of a campaign.
Yet polling, taken for granted today by politicians, pundits, and public, is a relatively new phenomenon. For most of American history, without an established statistical discipline or advanced communications, candidates had little way to know their likelihood of victory.
That wasn't for lack of trying. Americans have wanted to know who would win each race ever since campaigns became hotly contested affairs and engineered a range of original methods to do so. But creativity, of course, doesn't always translate into credibility.
The earliest polling methods appeared with America's first truly contested election in 1824. Andrew Jackson's supporters organized "test votes," or surveys of individuals, at ostensibly nonpartisan meetings. ("Nonpartisan meetings" included such gatherings as militia musters, which heavily favored the decorated general). The Jacksonians then disseminated their findings among publications.
Unsurprisingly, the only newspapers that considered these early "straw polls" credible were those that favored Jackson anyway. The Raleigh Register, which supported a rival, complained about "prematurely collecting the opinion of the people," writes historian Thomas B. Littlewood in Calling Elections: The History of Horse-Race Journalism.
Despite partisan misgivings, the desire to quantify public opinion spread. By the mid-19th century, taverns hosted spur-of-the-moment sample votes; coffeehouses displayed books that customers could sign to publicize their preferences. One poll was taken aboard a Mississippi steamboat, another at a North Carolina auction. Grand juries would survey themselves after dismissal; guests' toasts to candidates at parties would be counted.
Even cigars symbolized partiality. Tobacco stores would sell cigars from boxes affiliated with the various candidates, keep a running tally of how many of each were sold, and post it on their window. "These are crude, inaccurate measures of public opinion," says Kathleen Hall Jamieson, professor of communication at the University of Pennsylvania and an expert on campaign history. "You didn't actually know whether you were going to win or lose."
The result? In determining what the public wanted, candidates had to rely not only on instinct, but on information from local leaders. That became problematic by the end of the 19th century, as political corruption spread. To figure out what the public wanted, it became even more important to speak to them directly.
This was the approach that newspapers, joining the fray, took by the time of the 1896 election. Reporters polled people on trains, ships, even theaters. (This led to some strange findings, such as in 1908, the assertion that while William Howard Taft's devotees mainly patronized the upper-class theatres, burlesque audiences favored his rival).
Many realized the limitations of their methods and tried to fix them, says Kathleen Frankovic, director of surveys at CBS News. The Chicago Record, for example, sent postcards to all of the city's 328,000 registered voters for them to return with their preference marked. Before putting it into print, however, it had mathematicians check the results over to ensure that they were credible.
Still, it wasn't until the 1930s that polling took on a gloss of scientific reliability. A magazine called Literary Digest predicted who would win races based on sample ballots that its readers mailed in. "They're really the ones who put the practice of polling on the map," says Edward O'Donnell, coauthor of Visions of America: History of the United States. "They nailed a couple of elections almost perfectly."
The problem, however, was that the Digest depended on the sheer numbers of readers' responses, not on whether these readers represented the population at large. And so, in 1936, based on sample votes from 2 million readers, the magazine predicted a landslide victory for Republican Alf Landon. Franklin Roosevelt won with an Electoral College count of 523 to 8.
Corrected on 10/29/08: An earlier version misspelled the name of Kathleen Frankovic.