Time magazine poll: More proof of its bias

[WI_DEM]

I found some statistics regarding TIME magazines latest poll (the one which gives Bush a 11-point lead) Guess what? When respondents were asked who they voted for in 2000, those who had voted for Bush are vastly over represented.

52% of respondents in Times poll said they voted for Bush while only 41% said they voted for Gore.

Now it seems to me it was a much closer election than that in 2000 and that Gore actually won the popular vote. Interesting, huh?

See, pollsters can get any result they want to fit their biases.

http://www.srbi.com/time_poll.html

click on survey Questionaire
and check out page 20.

[RobertSeattle]

1. Great observation

I'm no polling guru, but I do like polls that try to "baseline" the the poll with "known value" responses.

[KurtNYC]

2. Propaganda with the goal of manipulating the election

Check their sister entity AOL -- for ten days they have had this stupid "President Match" thing in the middle of the page. Notice it doesn't say "Candidate Match" and the picture has been B*sh the whole time with his would-be Jesus hand outstretched.

http://www.aol.com

Blantant bias but it pretends to be just a tool to help you pick the pResident (as in "pick B*sh"). Obviously it is aimed at undecideds. At what point does stuff like this constitute illegal (undecalred) campaign contributions?

[NewYorkerfromMass]

3. Fraud! Forgery! Boycott TIME!!!!

:)

[DaveinMD]

4. remember

sometimes people will not admit they voted for the candidate that lost. One year after the 1960 election, over 60 percent of the nation claimed they voted for Kennedy. That was a 50-50 election as well.

[Jim__]

6. If your point is that people aren't answering the poll honestly ...

... well, that sort of makes the point that the poll is worthless.

[Jim__]

5. I don't think the dates in the questionaire were current

53% voted for Bush in the Aug 24-26 survey. I don't think we know the break down for the survey taken Sep 7 - 9. Bit, I agree, they appear to have over-sampled republicans. Their methodology is quite interesting, They say:

Weighting
Once data collection is complete, we check to make sure it is representative of the whole United States population. Often data is adjusted so that it accurately reflects the whole population. Final figures are matched against U.S. Census Bureau breakdowns on age, sex, race, education, and region of the country. Data is also adjusted to account for the fact that people who share a phone with
others are less likely to be contacted than people who live alone and have their own phones, and that households with more than one telephone number have more chances to be called than households with only one phone number.


They are weighting individuals based on national population averages. It would be interesting to know what weights were assigned to each group; and are they basing their weighting on national averages. If so, then we'd need to work back through the answers by state to see if certain groups within certain states might be over-weighted - e.g. if they under sample African Americans nationwide; but over-sample them in, say North Dakota. This type of error could appear in poll after poll. I'm sure a good statistician could adjust for this; but, nothing in their methodolgy says that they are.

Their own write up on margin of eror recognizes that MOE within sub-groups is larger than MOE for the whole sample. Again, they can smooth this out; but their write-up on methodolgy doesn't say that they are.