WASHINGTON - In the summer of 1932, a young advertising executive in New York decided to put his novel market research methods to the test to learn whether Iowans would elect his mother-in-law as their state's first female secretary of state.
Because no Democrat had been elected to a high office in Iowa since the Civil War, few people took much notice when George Gallup predicted victory for Ola Babcock Miller - until she won.
It was not exactly the birth of modern polling in politics - that took almost 30 more years - but Gallup's Iowa poll was a landmark, the first scientific political survey.
Polling has become such an integral part of the contemporary political scene that it is difficult to imagine how American politicians did without it. Polls now not only try predict the winners in political contests but shape the issues of a campaign, determine which voters and regions candidates target, even determine the slogans they use.
In fact, politicians have never really done without them, though the methodology of early efforts was suspect before Gallup and two other pioneers, Elmo Roper and Archibald Crossley, came along.
As far back as 1828, the minions of Andrew Jackson were being dispatched to key states to assess Old Hickory's support among "men of political knowledge and integrity." Riverboat passengers would take the pulse of fellow travelers and mail in their results to the newspapers. Curious citizens would survey townspeople at "militia musters," as their neighbors marched through the town square.
"Everyone wanted to know who was going to win the election, more so then than now because this was such a great mystery," says Tom W. Smith, director of the National Opinion Research Center at the University of Chicago. "There were no national polls."
It wasn't until the late 1930s - during the presidency of Franklin Delano Roosevelt - that a president systematically used polling to gauge public opinion. . And it was not be until John F. Kennedy in 1960 that a presidential candidate used a pollster to help shape his campaign.
But politicians have never wanted to fly blind. "Politicians have always had an interest in where public opinion was," says Karlyn Bowman, who studies polling and public opinion at the American Enterprise Institute in Washington.
Today's practitioners have little doubt that the art of "scientific polling" is light years ahead of its rudimentary precursors in accuracy, speed and the sophistication of the questions.
"Scientific polling is built on the scientific principals of probability," says Frank Newport, the editor of the Gallup Poll in Princeton, N.J. "Convenient sampling has no scientific basis. Now, that doesn't mean that you can't go onto a street corner, find a hundred people and be right, but it would just be luck."
Maybe so, but it seems that Gallup's ancestors could get pretty lucky. In the current issue of the journal Political Science, Samuel Kernell, a political scientist at the University of California at San Diego, took a close look at Andrew Jackson's 1828 efforts to gauge his popularity in Ohio, a key swing state then as it is now.
Jackson's operatives surveyed civic leaders in each of the state's 14 congressional districts, concluding that Jackson would defeat President John Quincy Adams by 7,150 votes, taking 52.7 percent of the total.
Four months later, Jackson won by 4,143 votes, 51.2 percent of the ballots cast.
Jump to September of 1992, when the "scientists" polling for CBS and the New York Times predicted that Arkansas Gov. Bill Clinton would squeak by President George Bush, 38.2 percent to 36.3 percent in Ohio. Clinton did beat Bush, by all of 159 votes. Kernell concluded that the modern pollsters were "moderately superior" in vote counting, though they fell short of their 1828 counterparts in predicting district-by-district variations.
"Overall, however, these differences are less striking than are their similarities," Kernell said.
By the 1850s, political operatives were replaced by newspaper straw polls. Publications stationed ballot boxes throughout their circulation areas, then printed sample ballots to be stuffed into the boxes. Rather than using small but carefully chosen random samples, the straw polls relied on large, "haphazardly random" samples, Bowman says, and they weren't too bad at it. "Some of them were quite accurate, almost despite of themselves."
In the early 20th century, the popular magazine, Literary Digest, took straw polling to a new level, querying not only its own significant readership but lists of names gleaned from car registrations, telephone listings and other public records. Millions of letters went out. Tens of thousands came back.
And the magazine was almost always right - until 1936, when the editors confidently predicted that Kansas Gov. Alf Landon would soundly defeat Franklin Roosevelt. No doubt, Bowman said, among telephone subscribers, car owners, and magazine readers, Roosevelt was deeply unpopular, but this was the Depression, and Roosevelt's partisans could not afford such luxuries.
Gallup's fledgling American Institute of Public Opinion, established a year before the 1936 election, made no such mistake on the outcome, nor did two other polls, by Roper and Crossley. And Gallup had begun polling the public on remarkably familiar subjects with thoroughly contemporary results.
Gallup's first syndicated column, on Oct. 20, 1935, asked voters, "Do you think expenditures by the government for relief and recovery are too little, too great, or about right?" The results: 60 percent said too great, 31 percent said about right, and 9 percent said too little.
Roosevelt took notice. His war office began commissioning surveys on the public's support for entry into World War II, and, once the war began, on morale, savings bonds sales and other war issues. Even then, the president was aware of the public's sensitivities to a politician putting his finger to the wind. Bowman said correspondence has surfaced in which Roosevelt stressed to pollster Hadley Cantril that he could not be seen meeting with administration aides at the White House.
If straw polls suffered defeat with Landon in 1936, scientific polling was set back in 1948. So confident were the pollsters in New York Gov. Thomas Dewey's victory over Harry S. Truman that they stopped polling weeks before the election. When Truman beat Dewey by nearly five percentage points, newspapers canceled their subscriptions to Gallup and Roper.
In 1960, Lou Harris hitched onto Kennedy's campaign, polling on the public's concerns about education, on Kennedy's Roman Catholicism, even on competition with the Soviet Union.
Polling evolved from there. Harris did 66 polls for Kennedy, 26 from September to November, according to Bowman. Two decades later, pollster Pat Caddell conducted 133 state surveys for Jimmy Carter, 14 national polls and two market tests for his candidate's commercials.
Richard Wirthlin, Ronald Reagan's pollster, worked off a computer program tracking the preferences of 10,000 voters. In 1960, it took two weeks for pollsters to determine who Americans thought prevailed in the Kennedy- Nixon debates. In 1976, results came back by the next day. In 1996, opinion was pouring in 15 minutes after President Clinton and Sen. Bob Dole had completed their closing speeches.
In October 1996, before ballots had been cast in the Clinton-Dole race, pollsters were in the field, asking whether Vice President Al Gore and Republican vice presidential candidate Jack Kemp should have been the nominees of their parties.