Ever Since Dewey Won, Political Public Opinion Polls Have Proliferated

It was Sept. 7, 1948, and the Roper Organization had seen enough polls to predict flatly who would win the upcoming election for President of the United States: New York Gov. Thomas E. Dewey.

“The thought back then was that nothing ever happened after September, that everyone had their minds made up,” recalls pollster Burns W. Roper. “We said it’s all over, Dewey’s in, there’s no point polling.”

It didn’t work out that way. President Harry S. Truman surged to victory, and his surprise triumph fostered the growth of aggressive public opinion polling as a pervasive feature of national politics.

20 Network Polls

Four decades later, political polls are inescapable. The three major television networks have already conducted more than 20 presidential preference surveys this year, identifying putative leaders months before most voters knew who was even running.

That parade of polls will become a stampede as the 1988 race intensifies. There will be national polls, state polls, exit polls, tracking polls and the growing phenomenon of overnight polls.

And at the receiving end, experts say, will be millions of voters who remain largely misinformed about polls, and dubious of their results.

‘Pieces of Evidence’

“Polls are more misunderstood than understood,” said Andrew Kohut, president of the Gallup Organization. “They’re thought of as predictions and conclusions about public opinion. In fact, they’re just pieces of evidence.”

Kohut blames analysts who “over-interpret” poll findings. “The easy thing is to say the public ‘feels’ this way about this issue,” he said. “The real thing is to say the public responded this way to this question.”

That caution is required because public opinion polling is as much high art as hard science, a marriage of language and mathematics in which every nuance counts. Bias in a question can wildly skew the results.

Conflicting Polls

Take, for example, two apparently conflicting polls last summer on U.S. policy in the Persian Gulf. One found 57% support for the use of force, if necessary, “to ensure an adequate supply of oil from the Middle East.”

That question was abstract. The other poll was specific: Should the United States blockade, bomb or invade Iran, if Iran attacked an oil tanker sailing under the U.S. flag? Put that way, only 29% said yes.

No less a political veteran than President Reagan defined the issue when he was asked about polls. “You have to know what questions are being asked, and how they’re being asked,” he said.

Statistical Theory

If question wording is the art of polling, the science lies in selecting who is to be asked the question. Pollsters, using statistical theory, reach a random group whose characteristics, such as sex, age, race, income, religion and geographical distribution, mirror those of the nation.

Done correctly, the sample can be astonishingly precise. “People have a hard time believing that,” said Harry O’Neill, president of the National Council on Public Polls.

“I tell them, if you really don’t believe in sampling, the next time the doctor draws a blood sample, say, ‘No, Doc, take it all.’ ”

There are some mistakes--telephone surveys, for example, miss people who do not have phones--and pollsters tinker with their figures to adjust for such errors. They also use a mathematical formula to arrive at the sampling error, an estimate of how close the results would be if every adult in the nation were asked the same question.

More Misunderstanding

But sampling error, while important, is another area of misunderstanding. “It sucks people into thinking it’s the only source of error,” when in fact question wording is as important, O’Neill said.

A good sample means that a pollster has selected a representative segment of the population; it does not mean the pollster has asked those people valid questions.

Roper, who wants to lower the emphasis on sampling error, conducted a poll in which only 14% of the respondents correctly identified the definition of sampling error from among four choices. The largest group, 29%, thought sampling error meant “all possible sources of error.”

Inherent Imprecision

Sampling error remains important because it indicates the imprecision inherent in polling. If a poll’s sampling error is plus or minus 3%, a candidate reported to have 24% support in fact may have anywhere from 21% to 27% backing. If another candidate is reported to have 22% support, the two can only be said to be tied.

Such horse-race polls, pitting candidates against one another, garner the most attention in a political campaign.

“It is like wanting to know who wins the World Series before it’s over,” Roper said. “It’s human curiosity.”

But critics say pollsters have run amok in their effort to satisfy that curiosity. Already this year, “There’s been an awful lot of activity, more than I’ve noticed in a long time,” said I. A. Lewis, poll director for the Los Angeles Times. “I do more polls that I really think are necessary, simply because of the demand.”

Bush Leads Field

The polls this year have consistently put Vice President George Bush well ahead of the Republican field. And Gary Hart, before his withdrawal in May, led his little-known Democratic opponents. After his withdrawal, no other Democrat fared as well in the polls.

Before voters are attuned to the campaign, Lewis said, horse-race polls mainly show recognition of candidates’ names. In his view, “Almost all of the stories dealing with standings right now have almost no meaning at all.”

That can change, and quickly. Voter opinions can evolve so fast some pollsters do nightly surveys in the days before important primary and caucus elections, adding each new night’s result to the previous few nights’ totals to get a rolling figure that shows movements in support.

Gut Reactions

Another phenomenon, stand-alone overnight polls, gives fast readings of opinions on events. But such polls have greater chances of error, and some critics say they measure gut reactions rather than thoughtful opinions.

For instance, Hart immediately topped the Democratic race when he rejoined it in mid-December. But pollsters cautioned that an unusually large number of voters had strong negative feelings toward Hart.

If some of their work merely satisfies curiosity, pollsters say, they also can serve a more important function. “Polls keep the leadership of the country in closer touch with the views of the American public,” Kohut said. “And that isn’t bad in a democracy.”

Then again, there are skeptics. “A man who is influenced by the polls . . . is not a man to represent the welfare of the people,” wrote one.

His name was Harry Truman.