|
As a market research statistician, I can tell you most of these polls are a joke. Some of them, particularly the CBS ones favoured by the Clark and Kerry folk, are now being put out with a SIX PER CENT MARGIN OF ERROR! That's pathetic.
I mean, with a six per cent margin of error, you're saying:
"Kerry is somewhere between 2% or 14%."
"Clark will lose to Bush by 11% or beat him by 1%."
"Dean will lose to Bush by 12% or lose by 2%."
That degree of confidence in the polls is near-meaningless, and the sample sizes (under 1,000) are much too small.
None of the polls looking at candidates are serious. I mean, does anyone REALLY believe that John Kerry has less support amongst Democrats than Al Sharpton or Carol Moseley Braun? Yet that's exactly what some of these polls are saying.
Also, consider the HUGE disparities between polls being conducted simultaneously. One says that Clark and Dean are tied against Bush. Another says that Clark loses, Dean loses even more. Another shows Dean ahead of Clark by six percent and barely losing to Bush. All were taken at the same time. Which one is right?
Probably none of them.
There are two things to consider when looking at polls:
1) Sample size -- is it large AND representative of the voting population?
2) Margin of error/degree of confidence -- any poll with a larger margin of error than 2 to 3 per cent is utter junk that would get you flunked out of a uni statistics class.
Even better polling services get it wrong. Zogby showed Clinton and Dole "neck in neck" before the 1996 election (where Clinton pasted Dole). Gallup showed Bill Clinton at 25% to Bush's 60% at the time of the Democratic nomination (assuming Perot wasn't in the race). We all know how that went.
|