Ever want to know what to look for in reading poll results?????
http://www.nypostonline.com/news/37215.htm
Labor Day is when analysts say campaign polls really start to matter.
But not all polls are equally reliable, so here are a few tips on how to read them:
* Rely on polls of "likely voters" - polls of "registered voters" who are eligible to vote (but may not bother) or "adults" (who aren't all U.S. citizens) are cheaper to do but less reliable.
* Polls of "registered voters" and "adults" tend to show Democratic support higher than it will be on Election Day - because poor and less-educated voters tilt Democratic but are less likely to actually vote.
* Look at the sample size - the smaller it is, the less reliable the poll. Many analysts say a national poll should have a sample of at least 1,000 to be credible.
* Consider a poll like tracking a diet on a scale - like scales, you can't jump from poll to poll and be sure a trend is real. So if Poll A has a candidate ahead by one point and the next day Poll B has him up 3 points, it could be a variance between polls rather than a real trend.
* Or you can try the Poll of Polls approach - average all polls for a given period of time. A variation is to drop the poll that looks best for Gore and the one that's best for Bush, and average the others.
* Pay special attention to independents - particularly in a close race - because they're likely to decide the winner.
* Check the split between men and women. For the past few decades, men have usually tilted Republican, so the real question in 2000 seems to be where women go.
Remember that a majority of men even backed hapless Bob Dole over Bill Clinton in 1996. Any poll that shows men favor Al Gore over George W. Bush means Bush is dead - or the poll is dead wrong
* Past performance, as they say on Wall Street, is no guarantee of future results. But the polls with the best predictions of the 1996 vote were John Zogby's poll for Reuters and the "Battleground" poll by Republican Ed Goeas and Democrat Celinda Lake.
All this explains why many analysts are skeptical about the new Newsweek poll that put Gore 10 points ahead - it's "registered" rather than "likely" voters and has a small sample of under 1,000. Its 755 "registered" voters might include only 600 "likely" voters.
Also, Newsweek suggests Gore is breaking even among men, but almost every other poll this year has put Bush way ahead with men.
Newsweek senior editor David Alpern said the poll uses registered voters because the pollsters think it's "still too early" to switch to likely voters.
But if the Zogby, "Battleground" and Gallup/CNN polls due out later this week have similar results, it's a good bet that Bush is in big trouble.
------------------
Extremism in the defense of liberty is no vice. Moderation in the pursuit of justice is no virtue.
Barry Goldwater--1964
http://www.nypostonline.com/news/37215.htm
Labor Day is when analysts say campaign polls really start to matter.
But not all polls are equally reliable, so here are a few tips on how to read them:
* Rely on polls of "likely voters" - polls of "registered voters" who are eligible to vote (but may not bother) or "adults" (who aren't all U.S. citizens) are cheaper to do but less reliable.
* Polls of "registered voters" and "adults" tend to show Democratic support higher than it will be on Election Day - because poor and less-educated voters tilt Democratic but are less likely to actually vote.
* Look at the sample size - the smaller it is, the less reliable the poll. Many analysts say a national poll should have a sample of at least 1,000 to be credible.
* Consider a poll like tracking a diet on a scale - like scales, you can't jump from poll to poll and be sure a trend is real. So if Poll A has a candidate ahead by one point and the next day Poll B has him up 3 points, it could be a variance between polls rather than a real trend.
* Or you can try the Poll of Polls approach - average all polls for a given period of time. A variation is to drop the poll that looks best for Gore and the one that's best for Bush, and average the others.
* Pay special attention to independents - particularly in a close race - because they're likely to decide the winner.
* Check the split between men and women. For the past few decades, men have usually tilted Republican, so the real question in 2000 seems to be where women go.
Remember that a majority of men even backed hapless Bob Dole over Bill Clinton in 1996. Any poll that shows men favor Al Gore over George W. Bush means Bush is dead - or the poll is dead wrong
* Past performance, as they say on Wall Street, is no guarantee of future results. But the polls with the best predictions of the 1996 vote were John Zogby's poll for Reuters and the "Battleground" poll by Republican Ed Goeas and Democrat Celinda Lake.
All this explains why many analysts are skeptical about the new Newsweek poll that put Gore 10 points ahead - it's "registered" rather than "likely" voters and has a small sample of under 1,000. Its 755 "registered" voters might include only 600 "likely" voters.
Also, Newsweek suggests Gore is breaking even among men, but almost every other poll this year has put Bush way ahead with men.
Newsweek senior editor David Alpern said the poll uses registered voters because the pollsters think it's "still too early" to switch to likely voters.
But if the Zogby, "Battleground" and Gallup/CNN polls due out later this week have similar results, it's a good bet that Bush is in big trouble.
------------------
Extremism in the defense of liberty is no vice. Moderation in the pursuit of justice is no virtue.
Barry Goldwater--1964
