IN EACH OF the last five Presidential elections the major public opinion polls have predicted the winner and his margin of victory within a few percentage points. Consequently, today, the scientific sampling techniques used by most polling organizations have come to be accepted as a fairly reliable barometer of American public opinion. But there are major questions involving the uses--and misuses--to which the polls are put.
There are approximately 200 public and private polling organizations in the United States today; and most of the reputable ones have similar methods for conducting their surveys. The Gallup poll uses 300 sample precincts throughout the nation. These are chosen at random with the stipulation that the number in cities, towns, and rural areas be proportionate to the percentage of the population living in such areas. A Gallup interviewer, usually a housewife working part-time, is assigned to each precinct and told to start at a specific, randomly selected household and move from house to house until she has five interviews.
When each interviewer has five interviews, he mails them to the Gallup headquarters in Princeton, New Jersey. The data is processed and the results, based on a sample of 1500 persons, are released to the news media. From start to finish a poll takes roughly eight days--with the exception of the final one, published the day before elections, which is processed more hurriedly.
The basic error, called the "sampling error," is plus or minus two to three per cent; this is mathematically inherent in the small sample size. To decrease this error by half, the sample size would have to be quadrupled, which would be very expensive and would slow down the process so much that the results published would no longer be an accurate reflection of public opinion at the time.
Louis Harris, America's other leading public pollster, has added his own innovation to this technique. In each consecutive poll his organization interviews many of the same people over again in an effort to get a more consistent picture of shifting trends.
The main check on accuracy is competition. No one would consider faking a poll because it would immediately be challenged by the others. Pollsters want to be accurate--that's how they make their money.
THEIR VERY accuracy--at least in certain circumstances--makes the public opinion surveys an important political tool. Even as simple mirrors of public opinion they can have far-reaching side effects. For instance, no one knows how many voters last month were swayed by a tendency to jump aboard a Nixon band wagon. Leading pollsters, including Gallup, unequivocally reject the notion that there may be a so-called "band-wagon effect." They cite Hubert Humphrey's dramatic comeback as evidence for their view. Still others feel that the polls may actually have helped Humphrey by generating an "underdog" sympathy vote. Whichever of these effects was dominant, it seems obvious that in an election where only a few hundred thousand votes out of more than eighty million decided the outcome, the polls could have had a telling effect.
Financially, Humphrey felt the polls had hurt him. He remarked several times that he received fewer donations at the outset of his campaign because the polls had marked him with the stigma of defeat. A more direct danger, however, is that polls can be deliberately manipulated by office holders and office seekers alike to influence public opinion.
There have also been instances of politicians commissioning private polls of limited samples and leaking the results to the press as if it were a national poll. By adjusting the sample it is, of course, possible to obtain any desired results. In recent campaigns, major candidates have frequently commissioned polls on certain issues, using the data to mold a popular campaign image of themselves. This sort of molding is, obviously, what politicians have always done; but it may not be in the interest of better leadership that they have an instrument as fine as the polls to help them.
AS POLLS are used in more sophisticated ways, new problems arise. It is difficult to accurately determine responses to issues of a complicated nature. This October the New York Times commissioned the Gallup organization to do a poll on various aspects of living conditions in Harlem. There were some 86 questions of a rather personal nature and the interviewers had a great deal of trouble getting answers.
When the poll was completed, a Times editor sent a reporter to a few of the addresses polled to get some direct quotes and discovered that the buildings didn't exist. Gallup scrapped the poll when he was told, and explained that because only black interviewers could be used it had been necessary to hire some people who were not on the regular staff. Two of these had falsified their data. Gallup explained that one of the primary means of checking interviews--spot checks by telephone--had been ineffective because there are so few phones in Harlem. He didn't explain why two other habitual means of checking--postcards and the so-called "cheater questions"--had failed. The Harlem survey was the first of the 8000 polls Gallup has conducted to be scrapped.
Another problem that has emerged in the recent campaigns is the extreme volatility of voters. Three consecutive polls in the last ten days of July showed radically different estimates of the respective strengths of Nixon and Rockfeller against various Democratic opponents. Gallup and Harris, two of the three involved, issued a joint statement explaining the discrepancy on the grounds of a shifting political scene, high voter volatility, and sampling errors. They added that, "each [poll] was an accurate reflection of opinion at the time it was taken.
While this explanation seemed satisfactory, Rep. John Moss (D-Calif.) saw danger in the nation's two chief pollsters collaborating on a joint statement. He submitted a bill to the House of Representatives which provided for the establishment of a Committee on Public Opinion Polls to look into all aspects of the problem.
Also before Congress is a "Truth-in-Polling" Bill, submitted by Rep. Lucien Nedzi (D-Mich.). This would require that within 72 hours of publication of any political survey the following facts be filed with the Library of Congress and made available to the public:
* The person who commissioned the poll.
* The details of the sampling.
* The questions asked, the time period during which the interviews were conducted, and the method of interviewing.
* The results of the poll.
If passed, these bills would do much to ensure the accuracy and honesty of polling techniques. Nevertheless, the plethora of increasingly subtle problems, many of which have only emerged in the last year, would remain unresolved.-