How can one poll show 6 per cent undecided, and another poll the next day show 49 per cent undecided

[Skip Intro]

seems a little strange, doesn't it?

------------

Clemson University Palmetto Poll
11/21/2007

South Carolina
w/o Al Gore
Added: 11/29/07
Margin of Error = 3.6%

Hillary Clinton 19%
Barack Obama 17%
John Edwards 12%
Joe Biden 2%
Bill Richardson 1%
Unsure 49%

------------

Rasmussen Reports
11/20/2007

South Carolina
w/o Al Gore
Added: 11/22/07
Margin of Error = 4.8%

Hillary Clinton 43%
Barack Obama 33%
John Edwards 11%
Joe Biden 2%
Mike Gravel 2%
Bill Richardson 1%
Dennis Kucinich 1%
Unsure 6%

-----------

Survey USA
11/10/2007

South Carolina
w/o Al Gore
Added: 11/14/07
Margin of Error = 4.6%

Hillary Clinton 47%
Barack Obama 33%
John Edwards 10%
Unsure 5%
Other 5%
------------

http://www.usaelectionpolls.com/2008/polls/South-Carolina-November-2007.html

[Jackpine Radical]

1. I have a couple of theories about that.

but I'm not sure which is correct.

[eyesroll]

2. One poll might count leaners or press more heavily for an answer.

For instance, if someone answered "undecided," one poll might just take that and move on. Another poll might follow up and ask if they were leaning a particular way, or if they would stay home if the election was today.

[depakid]

3. Crappy methodology

Good research is expensive, which is why campaign spend so much on internals (and keep the results close to their chests).

Could also be agendas- certain polls are know to skew results to influence public opinion (as opposed to reflecting it).

[msongs]

4. the only valid poll is the one that validates one's own position :-) nt

[elizm]

5. Door to door canvassing in South Carolina....

I can tell you that 35-40 percent of Democrats ARE undecided.

[DemocratSinceBirth]

6. Some Pollsters Push Undecideds And Leaners To Make A Choice

Usually it's nothing nefarioius...

[Adelante]

7. It's not strange

It's why you want to compare a new poll with a former poll done by the same company and not by two companies who may have used different methodology. An aggregate of all polls is also okay to get an idea of positioning, but comparing the results of the different polls doesn't really tell you anything.