ad info

CNN.com
 MAIN PAGE
 WORLD
 ASIANOW
 U.S.
 U.S. LOCAL
 ALLPOLITICS
  TIME
  analysis
  community
 WEATHER
 BUSINESS
 SPORTS
 TECHNOLOGY
 NATURE
 ENTERTAINMENT
 BOOKS
 TRAVEL
 FOOD
 HEALTH
 STYLE
 IN-DEPTH

 custom news
 Headline News brief
 daily almanac
 CNN networks
 on-air transcripts
 news quiz

 CNN WEB SITES:
CNN Websites
 TIME INC. SITES:
 MORE SERVICES:
 video on demand
 video archive
 audio on demand
 news email services
 free email accounts
 desktop headlines
 pointcast
 pagenet

 DISCUSSION:
 message boards
 chat
 feedback

 SITE GUIDES:
 help
 contents
 search

 FASTER ACCESS:
 europe
 japan

 WEB SERVICES:
 TIME on politics Congressional Quarterly CNN/AllPolitics CNN/AllPolitics - Storypage, with TIME and Congressional Quarterly

How are CNN polls taken?

By Brooks Jackson/CNN

February 18, 1999
Web posted at: 5:23 p.m. EST (2223 GMT)

WASHINGTON (February 18) -- How is polling at CNN carried out?

Designing a poll begins with the questions. Experts at CNN and its polling partner, the Gallup Organization, first formulate the questions. While writing them, the two groups wrestle with the wording. Is it biased? Will people understand what's being asked?

RELATED VIDEO
CNN's Brooks Jackson takes a closer look at how polls are taken
Windows Media 28K 80K

Once the questions are settled, Gallup orders up a list of who will answer them. On average, Gallup completes just over 1,000 interviews to represent a nation of more than 270 million. But if it's a truly random sample, it's like stirring soup -- so that one spoonful will tell you what the whole pot is like.

This sample will be a list of telephone numbers generated by computer -- random digits, so even unlisted phones will be included. The calls go out from eight Gallup centers around the country.

Here's the main thing -- the phone in your home is just as likely to ring as any other in the whole country. The only exceptions: Hawaii and Alaska, because of the expense.

A computer dials the random numbers and interviewers often reach business, fax machine or computer lines.

Not every answer is counted equally. People with two home phone lines, for example, are twice as likely to be called, so their answers are weighted less heavily so two-phone households won't be over-represented.

Any poll has a margin of error. Mathematically, a truly random sample of 1,022 should, in 19 polls out of every 20, produce results that are within three percentage points of what we would come up with if we interviewed everybody with a telephone.

And of course everyone who does not have a telephone is missed. That's about 6 percent of the population, more often poor, Southern and African-American than the population as a whole.

But going door-to-door would be many times more expensive. And it's been a long time polling flunked the election test.

In 1996 CNN's last poll before election day showed Bill Clinton 52 percent, Bob Dole 41 percent and Ross Perot 7 percent. The actual vote: Clinton just under 50, Dole 42, Perot 8. Not bad.

And, next time we call, it could be your phone.


RELATED STORIES

Poll: Clinton scandal has not taught young Americans it's OK to lie (2-17-99)

Poll: Republican image hurt by impeachment process (2-15-99)

Poll: Clinton wins, GOP and public lose in impeachment saga (2-14-99)

Poll: Most people want Clinton censured but not convicted (2-10-99)


RELATED SITES

Gallup Organization Web site



MORE STORIES:

Thursday, February 18, 1999

Search CNN/AllPolitics
          Enter keyword(s)       go    help


© 1999 Cable News Network, Inc. All Rights Reserved.
Terms under which this service is provided to you.
Read our privacy guidelines.
Who we are.