« CNN says, She Says | Main | Laziness vs Ignorance »
September 29, 2004
Gallup
I don't want to necessarily pile on to Gallup on the party ID thing. It's hard to tell exactly how you reweight for party-id if at all. I've seen statewide private polling in Georgia that clearly had an improportional sample, for instance more Republicans in DeKalb County than Democrats. In that example, maybe reweighting isn't the remedy but you clearly take that into account when you're looking at the poll results.
Instead of attacking party ID weighting, my aim is to attack the screen. Here are the seven questions that Gallup asks.
- 1. How much have you thought about the upcoming election?
- 2. Do you know where your polling place is?
- 3. Have you ever voted at your precinct before?
- 4. How often do you vote, always, often, etc...
- 5. Do you plan to vote in the November Presidential Election?
- 6. In the last election, did you vote or did something come up which prevented you from voting?
- 7. Rate your intention of voting from 1-10
OK. Now imagine two voters. One is ex-military, has lived in Pierce County Georgia his entire life. He answers yes to 1,2,3,4,6 and 7. He's disgusted with the way things are going in Iraq, is disgusted with Bush but also thinks John Kerry is a communist traitor. So he just plans on skipping the election for President but will show up to vote in the US Senate race and for his friend running for county commissioner. After he gets through the screen (getting 6 of 7 questions right) he tells the pollster that he would choose Bush if he had to make a choice, but doesn't plan on voting in that race.
The other voter is a 25 year old African American woman who just moved to South DeKalb county from rural Alabama. She has voted once before, so she can answer yes to question #4, she's following the race closely and plans to vote so she can answer yes to 1,5 and 7 also. At the present time, she doesn't know where her precinct is (no to 2) and obviously hasn't voted there before (no to 3) and the first time she voted was in 2002 so she didn't vote for President last time (no to 6). However, she absolutely intends to vote for John Kerry this time, and a friend told her about early voting so she'll just go to the county election office on Memorial Drive (next to the jail) instead of finding out where her precinct is.
So, voter #1 got a score of 86% on the likely voter screen, and voter #2 got a score of just 57%. Voter #1 is considered more of a likely voter than voter #2 even though when asked if he was planning to vote in this election he said NO.
This may have been a good way to do it in 1948 when mobility wasn't as high as it is now. But the idea that because 55% of the population votes you can just take the top 55% in some arbitrary 7 question survey should raise serious red flags -- especially for someone who is an expert in statistics.
Posted by Chris at September 29, 2004 12:57 PM
Comments
Post a comment
Thanks for signing in, . Now you can comment. (sign out)
(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)