Bookworm on Sep 22 2008 at 11:40 am | Filed under: Media matters
Related posts:
Email This Post To A Friend
You must be logged in to post a comment.
Earl on Double paying in Britain for health care
Oldflyer on No matter how you read the Second Amendment, it comes up favoring guns in citizens’ hands
Charles Martel on Fisking three dishonest Democrat senators on the subject of ObamaCare’s birth control mandate
Indigo Red on No matter how you read the Second Amendment, it comes up favoring guns in citizens’ hands
Bookworm on Fisking three dishonest Democrat senators on the subject of ObamaCare’s birth control mandateBookworm on William Shatner on gun controlBookworm on Using the First Amendment to nullify God — Air Force editionBookworm on No matter how you read the Second Amendment, it comes up favoring guns in citizens’ hands
jj on Double paying in Britain for health care
Can anyone find anything *wrong* with the following analysis of the recent Gallup polls?
The conclusion of the analysis is: In a week where McCain gained strength and Obama lost strength, the Gallup result said a four point swing in favor of Obama. Why? SOLELY because they drastically modified the weight % of Dems vs Repubs, to reflect not a McCain uptick, but an Obama uptick.
The evidence looks pretty damning to me. Am I missing *anything* else that would explain it, or missing anything that would refute the analysis?
http://wizbangblog.com/content/2008/09/19/how-liberal-trolls-are-working-to-get-mccain-elected-president.php
Yes. The conclusion presupposes that Gallup deliberately skewed the sample to get a particular result, ie load it up with groups known to favor one side or the other. That’s unlikely. They probably screen for, and have a design quota, for affiliation (D,R,I) but are not likely to screen for answers to demo questions: “would you describe yourself as a:
Liberal Democrat
Moderate Democrat
Conservative Democrat
Independent
Liberal/Moderate Republican
Conservative Republican ”
The percent distribution is more likely than not the luck of the draw. (The make-up of the sample).
That doesn’t mean that advocates can’t make the most of what the results show. I happen to think that is why the Media and a lot of the people are so shocked when the election turns out different from what they were expecting. I also have a strong suspicion that some percent of the responders lie to polsters for fun. (I just hang up.)
Ellie2, unless you are Sarah Palin’s mother, I’d say it is odd that you would think that campaign or press “attacks” on a single individual are a greater social ill than racism. And if you would welcome racism to defeat a candidate that you oppose for legitimate ideological or policy reasons, then you are a utilitarian of greater proportions than most socialists, perhaps including Alinksy. Why stop at accepting racism? Why not call for voter intimidation and fraud?
By the way, no one is requiring you to make certain judgments about people based upon skin color, only asking that you try to avoid making false ones based solely on skin color.
Sorry about the prior post, but I put it in the wrong spot…
Ellie2,
I don’t think the article suggested that the people sampled were deliberately chosen to skew the results. (I think that was your explanation, Ellie, right?)
If I understand the article itself, the actual results of the polling are modified by weighting the results in one direction or another, by comparing the percents-in-each-category of the respondents against the “known percentages”, and changing the numbers thusly. Then the new numbers are what they report.
The article claims that as of the final week, Gallup MODIFIED the “known percentages”. Therefore, when the set of changes in the actual respondents are guaranteed to benefit McCain, because they changed the weightings, they reported results that instead indicated a swing to Obama.
I appreciate the answer, though I don’t think there was deliberate sample selection of specific people… Does anyone else see any other reason for the discrepancy?
I’m not claiming motive on the part of Gallup here. Gallup did not even report their weighting percentages, or any other information like that, so motive cannot be discerned either. If you agree that they modified their weights during the last two weeks, I’d sure love to see those weight values, and an explanation of why they changed them so dramatically!
As a simple example… suppose I gave this (simplistic example) poll:
A. Are you a Democrat or Republican?
B. Who are you voting for, McCain or Obama?
Suppose 100 people answered the phone with:
Republican/McCain : 55 people
Democrat/Obama : 45 people
Now suppose the “actual percentages” that my organization used on Sept 14th were: 50% Republican, 50% Democrat.
Then I would put out a report saying: “McCain leads Obama this week, opening up a 10 point gap, 55% to 45%.
—–
Now suppose I repeated the poll on Sept 21st and got these results:
Suppose 100 people answered the phone with:
Republican/McCain : 58 people
Democrat/Obama : 42 people
But this week, I changed my “actual percentages” to:
40% Republican
60% Democrat
That means that with my new weighting, 46.4 people chose McCain, and 50.4 people chose Obama.
As percents, 48% for McCain, and 52% for Obama.
Therefore even though McCain actually increased his lead by 6% over Obama in the actual respondents, what would I report:
Headline: “McCain support plummets! McCain now trails Obama by 2%.
Story: “This is an incredible, massive shift in voter sentiment,” said Obama spokesman BlahBlah Blovia. “Clearly John McCain has had a perfectly dreadful week. His blunders, including his choice of the woeful Sarah Palin, are now taking their toll.”
All this, even though McCain’s support actually increased. Based solely on modifying the “weights”. We used to call that cheating when I grew up, but I’ll wait for Gallup’s explanation.
mike
since we’re talking about polls, you also have to take into consideration that polls only tap people on their rolls and who have lan lines. obama has seen an unprecedented amount of support from new voters; most of whom are young, tech-savvy people who don’t have lan lines. there is no way to track their support; and by all accounts it is sizable.
peace
Dagon provided us a great service by way of that demonstration.
Dagon,
Certainly you’re right about polls relationship landlines and cell phones and voip.
And other factors that cause people to hang up or not answer the phones.
But the fact that Gallup’s six subcategories all showed increases for McCain, yet the summary numbers showed a drop, remain (apparently) inexplicable except for the hidden shift in their weights. No other explanation has appeared yet and Gallup sure isn’t telling us.