I’m sure that by now most people have seen today’s Gallup National Tracking Poll that shows Romney ahead of Obama 51-45. They had Romney up 50-46 yesterday. There were no cross-tabs to look at for today’s number, but there were cross-tabs for yesterday’s results. I want to take a closer look at the geographic cross-tabs and the weighting they give to each region.
Here is what Gallup is showing in their cross-tabs by geographic region.
. Obama Romney Margin
. East 52 48 O+4
. Midwest 52 48 O+4
. South 39 61 R+22
. West 53 47 O+6
Romney is only winning one region. He is winning the South and winning it big. I’ve already seen a lot of arguments about how he can’t possibly be winning by that big of a margin in the South, and the East can’t be that close, and so on. I don’t care for the purposes of what I’m doing here. It doesn’t matter. I just wanted to see what the weighting of the population is in order to get the top line outcome that Gallup got.
The first thing I did was pick an arbitrary number of voters. I chose 129,391,711. That’s the total of the votes received by Obama and McCain in 2008. The number doesn’t matter for this purpose. I could have used 4 or 10 billion and the result would have been the same.
I set up a spreadsheet based on the geographic cross-tabs supplied by Gallup. There was a 4% undecided vote in their top line totals based on the 50%-46% result, so I took 4% from each geographic region and called it undecided. I took the remaining votes and assigned them to Romney and Obama based on the cross-tab percentages. I assumed an equal 25% weighting to each of the geographic regions. The results look like this:
Notice the final result. It is 49% Romney and 47% for Obama and 4% Undecided. Obviously that’s different than what Gallup had in their top line results. It’s a better result for Obama.
The only explanation for this is that Gallup did not use an equal weighting for each geographic region, so the next step is to figure out what weighting they used. This isn’t rocket science. The only way for Gallup to end up with a better top line result for Romney would be for the weighting to be higher in the South geographic region. It’s the only region where he’s ahead.
I started removing a percentage of weighting for each of the Obama regions and adding the offsetting 3 percent to the South region until I got to a 50% to 46% Romney lead. The results look like this:
I hate screaming about poll biases and skewing, but this really smells funny. Why would Gallup assume a 10% greater weighting in the South than all of the other geographic regions? I would like to hear a rational explanation if there is one.
The Presidential election doesn’t feel like a Romney +4 race right now, and it certainly doesn’t feel like a Romney +6 like Gallup had today. There are enough other polls that dispute Gallup’s findings. I just don’t understand why a reputable polling outfit would weight the regions this way to cause a result that makes their poll an outlier. It doesn’t make any sense.
7:12 PM PT: For the sake of argument I ran the actual numbers for 2008 using the same cross-tabs Gallup provided for that year. In order to get the actual result that occurred in 2008, using their geographic cross-tabs, the weighting would have been 26.41% for the South and 24.53% for the other three regions.
Did that many people move to the South since 2008?
Their poll may be completely legitimate. I'm just trying to understand why they have suddenly become the outlier.