I find there to be something incredibly wrong with this
CNN/USA Today/Gallup poll. Read through these first few paragraphs and see if you spot it.
Support for President Bush's proposal to revamp Social Security -- allowing younger workers to invest part of their Social Security taxes in private retirement accounts -- is sliding, according to a poll out Tuesday.
In the CNN/USA Today/Gallup survey of 909 adult Americans taken Friday through Sunday, 40 percent approved of President Bush's approach to Social Security and 53 percent disapproved. The question had a margin of error of plus-or-minus 3 percentage points.
When the polling question did not mention cutting benefits, Bush's proposal drew 45 percent support and 47 percent opposition. It was posed to 466 people, carrying a 4.5 percentage point margin of error.
But when 443 of the 909 polled were asked whether they supported private retirement accounts in exchange for a reduction of guaranteed retirement benefits, support fell to 33 percent, while opposition rose to 59 percent. The question had a margin of error of 4.5 percentage points.
So...if I'm reading this correctly, they polled 909 people. When they didn't mention benefit cuts, it was a nearly even split - 45% pro and 47% con. When they did mention any benefit cuts; the numbers fell dramatically, with only support by 33%.
So, then when they came out with the full numbers, for some reason they grouped together the 2 groups which were given different information and counted the whole sample as the whole poll. In other words, they asked 2 different questions to people and then put them in the same overall group.
On top of that, I'm not at all clear about how they came up with the %ages who were given each information. They asked slightly fewer than half of their sample the question including the benefit cuts, and slightly more than half of their sample the question without mentionning benefit cuts. Do they have data somewhere which says that 50% of Americans know that Social Security reform would mean benefit cuts and 50% don't? Why should I believe this poll any more than 1 which told 80% of its respondents that there would be associated benefit cuts. Presumably, that one would show about a 35% level of support for the plan, instead of the 40% number presented here. So, why should I believe either of those numbers, given that they've built that assumption into them?
I ask again, how is this a valid polling method?