Back at the Old Stand

The day after the 1948 election, Editor Henry P. Slane of the daily Peoria Journal (circ. 68,000) sent Pollster George Gallup a bristling telegram: CANCEL OUR SUBSCRIPTION. Like Gallup, Elmo Roper, Archibald Crossley and all the pollsters who had confidently predicted a Republican victory, Editor Slane had a morning-after headache. With the editors of some 30 other U.S. dailies who canceled their subscriptions to the polls, Editor Slane cried: "Never again!" But like many another swearing-off, it didn't take.

"I hate to admit it," said Editor Slane last week, "but we are using Gallup again." So were many of the other papers which had dropped him in 1948 and later, although Gallup does not have as many papers now as he did then (206 v. 226). Elmo Roper is not doing quite so well in the press either (only 54 papers v. 66 then), but his Sunday broadcasts over NBC are now carried on 90 radio stations v. 75 on CBS in 1948. Crossley says most of his 1948 clients are back.

Said the St. Louis Globe-Democrat's President James C. Burkham last week: "We took Gallup again only after Dr. Gallup himself came out to sell us. He proved to us singlehanded that he had changed his technique ... so he would never stick his neck and ours out again."

How have the necks been protected?

For one thing, right after the 1948 fiasco, the pollsters turned over their figures for an autopsy by the Social Science Research Council. It checked actual votes against the predictions. The council discovered three major errors:

P:The pollsters stopped polling too soon, thus missed the last-minute swing to Truman. One out of seven voters checked after the election said he had not made up his mind until the last moment.

P:The polls did not interpret the "undecided" vote correctly. Gallup had assumed the undecided would divide about equally between the two candidates; actually, 74% voted for Truman.

P:Polltakers erred in their sampling (e.g., they interviewed a higher percentage of college graduates than the actual population contains, a lower percentage of people with grade-school educations).

To correct such errors, Gallup intends this year to poll right up to the eve of election. He is also giving more attention to working out a method of determining how the undecided vote is likely to go. The undecided are being divided into groups on the basis of their attitude on key issues, how they have voted in the past, etc. On the eve of election, Gallup will allocate them to the two parties in accordance with the percentages indicated by these breakdowns.

Gallup is also increasing the polling among the lower income groups by assigning districts to polltakers. In 1948, when polltakers could pick their own areas, they often passed up slum districts. To this Gallup has added greater use of a check which market researchers have found accurate: "pinpoint sampling," in which every third house is visited in 150 U.S. electoral precincts picked at random. This is the method Gallup will use for his last pre-election poll, taken the Saturday before Election Day.

So far, the new methods have worked well. In1950's off-year election, Gallup was able to narrow his margin of error (average: 3.4%) down to less than 1%: the last poll indicated 51% for the Democrats, and they got 50.3%. In this year's New Hampshire primary, the Crossley poll, using similar techniques, gave Eisenhower a 51.7% majority; he got 52.3%.

Roper, who had quit polling two months before the election in 1948, did not try to predict the New Hampshire outcome. This year, in addition to reporting percentages, he is also trying to determine the issues or the qualities in a candidate which cause a voter to make up his mind. Said Roper: "We're not going to be just a fever thermometer." In Roper's poll this week, he found that only 27% are sure voters for Ike, only 23% sure for Stevenson. "The other 50% are either finding it pretty tough to make up their minds, or appear to have made them up with such strong misgivings as to indicate the definite possibility of a change later."

Crossley shows that Ike would win (52.8% v. 46.6% for Stevenson) if the election were held today. But Crossley's figures are based on the assumption that only 56% of the eligible Democrats will vote v. 72% of the Republicans. He concedes, however, that if everybody who is eligible to vote did so, Stevenson would win by a narrow margin.

The pollsters are well aware of possible error in their figures, but feel they have the alarm signals set. If they fail, it is still true, as Gallup says, that "nobody has yet devised a better way."

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