Roanoke Times
Copyright (c) 1995, Landmark Communications, Inc.
DATE: SUNDAY, January 12, 1992 TAG: 9201130236
SECTION: EDITORIAL PAGE: B-3 EDITION: METRO
SOURCE: BOB WILLIS ASSOCIATE EDITOR
DATELINE: LENGTH: Long
GALLUP CALLING
THE LATE George F. Gallup liked to say that a person's chances of being interviewed by one of his poll takers were about the same as being struck by lightning.One evening a couple of weeks ago, a bolt from the blue hit my left ear. "This is the Gallup Poll calling."
I was at home on a week night, not all that eager to be interviewed. It would take only about 10 minutes, the woman's voice said reassuringly. And when her first question was, "Do you approve or disapprove of the way George Bush is handling his job as president?", I thought I was about to see my own political views writ large on a national opinion-surveyor's tablet.
That was my second surprise. The Gallup organization, famous nearly 60 years for taking political polls, wanted no more of those opinions from me that night.
The approval-disapproval query, I learned later, is standard. But mine would be an "omnibus" survey, diverging into several other channels - some of them carrying their own surprises.
Which titles from a long list of financial-type magazines did I know? Which did I read? Which would I use for planning personal finances? Had I smoked in the past week? Had I dated co-workers? Made New Year's resolutions?
Presumably Gallup was, for one thing, conducting a survey on behalf of magazine publishers. And Gallup might want to know my religious background for demographic purposes, along with data on family income, education, etc. But why ask how often I went to worship services, or what I believed about the Eucharist?
Why, at another point, did Gallup's caller want to know what movie kiss I thought was most memorable? Or what my age range was when I received my first romantic kiss, and where it happened?
When I called Leah Stoker of the Gallup press office in Princeton, N.J., she had few ready answers. Gallup's staffers may not know all the reasons certain questions are asked. Sometimes the company is hired to gather information and never knows the results; they're proprietary data.
A subsidiary of Gallup is the Princeton Religion Center; maybe that's why some of the questions dealt with church. As for kissing, she theorized that some manufacturer is contemplating an advertising campaign that would focus on a particular group of the population.
Whatever. In Lincoln, Neb., where two of Gallup's telephone banks are located, Jane Miller, vice president of interviewing, said that because of the Gallup name and the care taken in selecting interviewers, "fewer people hang up on Gallup than on any other company."
Maybe a poll told them that. Certainly my interviewer, who said she was calling from Houston, Texas, had been both affable and professional. According to Miller, 15 people are turned away for every one hired to make these calls; Gallup has about 700 such interviewers in Houston, Lincoln and Omaha, plus a handful in Canada.
Gallup still knocks on doors, but 90 percent of its questioning is done by telephone: About 93 percent of America's 93.3 million households have one or more phones. Rita O'Donnell, a Gallup methodologist, explains that a data base of listed telephone numbers for the continental United States is divided into four regions. These are broken down into cities of 1 million or more people; cities of 250,000 to 999,999; and the rest of the country. In this end of Virginia, that last includes us.
From these divisions, Gallup draws "seed" telephone numbers: area code, three-digit prefix and two more digits. A computer randomly picks the last two digits. This procedure Gallup calls "a random digit stratified probability design." Gallup dials the number at least three times in its efforts to reach the party. For one opinion sample, calls are made to about 1,000 respondents across the country.
So Gallup has gathered my answers and those of 999 other people. Are these simply added up and then chopped into pieces, as for a pie chart? No. Each respondent's answers are assigned a "weight." The purpose is to reflect more accurately the person's views as a member of his/her demographic group, so that the breakdown of opinion won't be skewed because the sample happens to contain a disproportionate number of Wall Street brokers or out-of-work cosmonauts.
Gallup, used to handling these sorts of numbers, seems more comfortable with all this weighting than I might be. "The procedures," says the Gallup Poll Monthly, " . . . are designed to produce samples approximating the adult civil population (18 and older) living in private households (that is, excluding those in prisons, hospitals, hotels, religious and educational institutions and those living on reservations or military bases) - and in the case of telephone surveys, households with access to a telephone.
"Survey percentages may be applied to census estimates of the size of these populations to project percentages into numbers of people. The manner in which the sample is drawn also produces a sample which approximates the distribution of private households in the United States; therefore, survey results can also be projected to numbers of households."
Poll takers are confident enough in these methods that they are wont to say the findings are subject to error of, oh, plus or minus 5 percent. Smart-mouths may ask, if they're so sure about the size mistake they can make, why can't they just eliminate it?
Given an opinion sample large enough - say, all the adults in the United States - they could. But that wouldn't be polling. It wouldn't even be an election, inasmuch as not all adults ever vote.
Anyway, an election merely ratifies a set of temporary political decisions. Between elections, people's opinions keep changing. That's one of the things pollsters measure. Because they deal with shifting sands and changing winds, all they can do is try to be mostly accurate for a little while. They'll soon have it to do over.
When I asked, Gallup's folks didn't know how likely they are to poll you. One thousand calls at a time isn't many. But Gallup surveys more than a million people a year. Given that the nation has far fewer than 180 million adults - well, you too could be struck by lightning.