Oregon poll (Kerry: 38%, Bush: 34% or Bush: 44%, Kerry: 39%)

http://www.kgw.com/news/pdf/maypoll_complete.pdf

Presidential Debate

While consumer households in Oregon support John Kerry over George Bush (38% vs. 34%), likely voters tell us they would vote the opposite: George Bush 44% vs. 39% for John Kerry.

And I can tell you that the methodology here is very strange.

The researchers say they OVERSAMPLED REPUBLICANS for their "likely voter" pool of respondents, in order to be MORE REPRESENTATIVE. What in the world could this mean, except that they wanted to skew it? If you are using proper sampling techniques in the first place, your sample should be representative! This is not explained.

The second issue is that the confidence level is low, and the margin of error is high, which means that in the likely voter poll, there could be a statistical "dead heat" or a slight lead for one or both. This is kind of amazing, given they've said they "oversampled" Republicans!

http://www.wgoeshome.com

But it seems that they drew down the Republican support plus aligned the gender more towards women (57-43). Both of these moves would seem to bolster the Democratic response, yet Bush still leads. That was the most troubling part of the poll IMO.

Are there ways that doing this could favor Bush at the expense of Kerry? I guess if the Republican response was weak and they cherry-picked the Bush supporters, but the poll is pretty much in line with other polls as far as support by party is concerned.

but it's a valid, and very basic, research method. Say you sell widgets and you want to know how well a new product will sell. Say you know from previous results that 90% of your widgets are bought by men.

You poll. 60% of those polled like your product. Great, right? Nope, because your sample breaks down to: 50% women and 50% men, and while 80% of women like your new product only 40% of men like it. To find out what the real sales result will be, you have to weight the respondants by gender, which means that while you get a healthy 80% of your 10% female sales base (8% total), you only get 40% of your 90% male sales base (36%). The result is that only 40% of your total likely buyers like your product, so you cancel it.

:wtf: