November 2, 2000

Campaign 2000: Don't Be Fooled By The Polls

By Jeff Caldwell

KSU News Service

It is fun to follow poll results on television or radio, but donít count on the results being completely accurate.

According to Dallas Johnson, head of the Kansas State University statistics department, the margin of error is
critical in every poll. And in a close race, like that between Al Gore and George W. Bush, the margin of error can
give a false impression of the potential winner and loser.

"The problem with big polls, like ones from Gallup, is that they are national, and the elections are determined by
the electoral college." Johnson said. "They will get a popular vote as a whole, but there arenít enough people in a
given sample to say anything about a particular state."

According to Johnson, in a nationwide poll, such as many of those by the Gallup organization, steps are taken to
insure the most accurate possible figures. One way this is done is through a stratified random sampling technique.
This system examines the characteristics, such as age, sex and race, of the population of a certain region and
polls a sample of approximately 1,000 people that corresponds with the overall population as closely as possible.

This system of stratification also makes gathering information easier for the polling organization. "Because of
the stratifying, it allows you to get samples more economically than it would if you were to do it completely at
random," Johnson said.

Johnson said the posters stratify on the area code of the phone number for persons being called. After they select
an area code at random, they use random digit dialing to contact someone at that phone number.

"They try to conduct their polls on Monday through Thursday nights, as they know they get quite different
results if they call on Friday nights or the weekends," he said.