I’ve been watching the polls only a little bit, and in truth they don’t matter much – Obama has a lead, and we want our side to stay energized, so maybe it’s OK if his lead is portrayed as small in the media. (Then again, I’d love for the GOP to stay very discouraged for next couple weeks).
Still, something puzzles me...
Gallup, as you all know, is now publishing 3 sets of numbers every day. They publish their "all registered voters" numbers, and two different likely voter numbers. One uses their "traditional" likely voter screen, which is supposedly a little tighter (i.e., assumes younger voters don’t really vote); and they’re publishing a newer screen that assumes better turnout.
For a couple days, if you crunch the numbers, you can see that their models have basically assumed that nearly all the voters who get tossed for their traditional likely voter screen are Obama supporters. Today, they take it one step further If you go to their website, you’ll see all 3 numbers:
RVs: (2798) 52% Obama, 42% McCain
LVs new: (2277), 51% Obama, 44% McCain
LVs trad: (2590), 49% Obama, 46% McCain.
Weirdness #1: They have more people in the traditional LV screen than the new one.
Weirdness #2: Multiply the %ages by the #s and you get the following:
RVs: (2798) 1483 Obama, 1175 McCain
LVs new: (2277), 1161 Obama, 1002 McCain
LVs trad: (2590), 1269 Obama, 1191 McCain
So, their tight LV screen takes away 214 votes from Obama, but ADDS 16 votes to McCain’s total.
Given that the traditional/ tight LV screen is listed as having more people than the newer one, I’m assuming there is something wrong with the total. Still, for a company that gets the most press for producing their numbers, they ought to get their numbers right, dontcha think?
Anyone have a better explanation for these numbers?