POLLS: Confusing and Exaggerated

POLITICAL poll taking in recent years has proved considerably more reliable than oneiromancy, haruspication or the casting of rune stones. George Gallup has erred in his election polls by only 1.5% since 1954. Yet last week, Gallup and Louis Harris published such divergent interpretations of the preconvention G.O.P. standings that Gallup's son and partner, George Jr., admitted that the psephologists were facing "a credibility-gap problem."

Gallup's survey showed Richard Nixon defeating both Hubert Humphrey and Eugene McCarthy for the presidency. If the election took place at the time of the poll, Nixon would win, Gallup said, by two points over Humphrey and by five points over McCarthy. The sampling suggested that Nelson Rockefeller would merely tie Humphrey and defeat McCarthy by one percentage point of the vote.

Rocky's camp countered with a commissioned sampling by another veteran pollster, Archibald Crossley, who had surveyed the nation's nine major industrial states and found that the New Yorker could easily coast past both the Democrats, while Nixon would tie with McCarthy and defeat Humphrey by only three points.

Chorus of Hoots. Then Louis Harris published an independent national survey showing Rockefeller six points ahead of both Humphrey and McCarthy. Nixon, said Harris, would lose to Humphrey by five points and to McCarthy by eight. The wide discrepancies in the samplings brought a chorus of hoots--especially, of course, from the Nixon camp.

The nation's major pollsters have been in contact with one another since last June, when they formed the National Committee on Published Polls to publicize standards for their opinion surveys. Last week, when their contradictory reports appeared, Harris called George Gallup Jr., whose famous father was traveling in Europe, and persuaded him to join in an unprecedented joint public statement. After consulting Crossley, they issued a complicated collective verdict. If their three polls were "plotted out sequentially, as though they were conducted by a single organization, using the same sampling techniques and the same question-asking techniques," they concluded, then 1) a Nixon-Humphrey race would be extremely close, "with Wallace perhaps holding the balance"; 2) "Rockefeller has now moved to an open lead over his possible Democratic opponents, Humphrey and McCarthy"; and 3) "the McCarthy vote has shown and continues to show the greatest amount of volatility among the four leading candidates."

Error Margin. To justify the apparent turnabout, Gallup suggested that the timing of the poll taking was crucial. Gallup's sampling was made between July 19 and 21, just after Dwight Eisenhower endorsed Nixon. Ike's announcement may have swung some sentiment to his former Vice President.

Crossley's poll was taken between July 21 and 26, and Harris' between July 26 and 29. Some analysts point out that as Harris was doing his sampling, Rockefeller's saturation advertising and personal campaign activity was reaching a peak, while Nixon was vacationing briefly and Humphrey was recovering from the flu.

An error margin of as much as two or three percentage points is routinely assumed as a hazard of the pollster's trade--but that could hardly account for the startling discrepancies in last week's results. All three pollsters used basically the same techniques, although they often differ in their philosophies of interpretation. Gallup, for example, believes that "our job begins and ends with the reporting of facts." Harris argues that survey results are meaningful only if they are digested and interpreted. Each pollster has his own methods. Harris likes to reinterview some one he has already talked with, figuring that he can thus detect changes in sentiment over periods of time. Gallup, on the other hand, argues that a second interview makes a voter selfconscious, so that he gives a less accurate reflection of public sentiment.

Sheer Volatility. Most pollsters agree that a sampling of something like 1,500 people yields a fairly comprehensive picture of national trends and opinions. For last week's sampling, Gallup used 1,156 interviews gathered from throughout the country, Harris 1,346 and Crossley 1,976. All but 219 of Harris' samplings were reinterviews.

The sheer volatility of voter sentiment this year probably contributed to the wide differences in the pollsters' findings. Though the primaries were over, some of the same factors still applied: a plethora of candidates, no established party tickets, and moiling confusion in both parties over the issues. Despite the pollsters' fatiloquent accuracy in past years, they have stumbled often in the past six months. Almost all, for example, have consistently underestimated Eugene McCarthy's considerable strength. Harris and Gallup have frequently differed in their preference surveys, though never so widely as in the preconvention week.

If anything, last week's consensus statement simply made matters worse. In persuading Gallup to endorse the apologia, Harris may have widened the trade's credibility gap to the dimensions of 1948, when virtually every opinion sampling was ushering New York's Thomas E. Dewey into the White House. Twenty years later, the memory of that year sends shudders down the spines of all pollsters. One pollster called last week's results "a fiasco." Another, Burns Roper, observed: "If this statement of 'open lead' for Rockefeller is construed by readers as being designed to influence the outcome of the Republican Convention, it will be most unfortunate, both for the political process and for the public-opinion polling profession."

The pollsters' main problem--which led in part to last week's statement--is that politicians and the press are constantly exaggerating the importance of the polls. Many analysts, including pollsters, are beginning to wonder whether "racehorse" surveys taken in the heated confusion of a presidential campaign are worthwhile, since the temperature of the electorate can vary by half a dozen points from one week to the next.

This file is automatically generated by a robot program, so reader's discretion is required.