There’s a new poll out today that shows Obama and McCain virtually tied in West Virginia: 41.33 to 41.67, respectively. Here’s what Jake “The Genius” Stump had to say about it in the Daily Mail:
The poll, however, randomly sampled nearly twice as many Democrats than Republicans yet it still shows McCain slightly favored here.
Conducted over a two-day period last week, the group sampled 344 Democrats, 182 Republicans and 74 independents.
Oh noes! The poll oversampled Democrats and McCain still came out on top. Newsflash: The reason they polled nearly twice as many Democrats as Republicans is because in West Virginia . . . drumroll please . . . there are nearly twice as many Democrats as Republicans!
Democrats remain the dominant voting force in West Virginia, adding 26,416 to their ranks since the 2006 general election to total 675,305. Republicans gained 10,467 voters during that time, and now number 353,437.
If you pull out your trusty calculator, you will see that 675,305 divided by 353,437 is 1.9—that’s nearly twice as many Democrats as Republicans. The breakdown of the polling sample closely mirrors the breakdown of registered voters in West Virginia.
I’m not saying the poll is accurate. I have no idea whether it’s accurate. You’d have to ask Nate Silver about that. I’m just saying the sample was a good representation of the West Virginia electorate, as far as party affiliation goes. So there’s no need to offer a caveat that “the poll, however, randomly sampled nearly twice as many Democrats than Republicans yet it still shows McCain slightly favored here”.
Nice Red…. One technical point from a survey research egghead:
Given Democrats outnumber Republicans 2 to 1 in West Virginia, they were not technically *oversampled* in this poll. Oversampling occurs when you select a disproportionate number of people from some category to ensure that you will have enough cases in the dataset to conduct a meaningful analysis. When you do that, you always statistically weight the data to account for the oversampling.
As you correctly note, this poll has proportionate sampling.
I think this election has set a new bar for inane and idiotic hair-splitting over not particularly useful data. I’d like to say this is all over in just a couple of weeks, but it’s not. This silly bullshit is the new cool.
It’s not really hair-splitting, it’s a fundamental misunderstanding of how polling works. I’ve heard some Republicans complaining that the polls in this election have a liberal bias because they poll more Democrats than Republicans, so seeing an article in the Daily Mail that implied the same thing made me feel the need to correct it. If you think it’s silly bullshit, feel free not to read my blog.