Cosmic Conservative

Rational Conservatism from the heart of America

Things just don't add up...

By CosmicConservative
So, reading today's polls, which are still trending positively in Romney's direction, I am struck with some fundamental mathematical problems. I am supposed to believe from the polls that Romney has about a four point lead in the national polls, but that he is about four points behind in several key battleground states, particularly Ohio, Iowa, Minnesota, Colorado, etc... Yes there are some polls that show Romney tied and even slightly ahead in some places, but the overall polling averages still favor Obama. So, anyway, I went to the trouble of looking for the best, latest individual polling information from each state, then I used current population figures for each state to create a national average of the polls. Here are the results: According to a mishmashed average of the most current national polling, Romney leads Obama by about 50% to 47%. But if you total up all the states and combine all their extrapolated populations, Obama leads Romney by 47% to 46%. So there is a discrepancy between the two of 4%. If you suppose the individual state polls are better data than the national polls, that would suggest a narrow electoral vote win for Obama. But if you suppose the national polls are better data than the individual state polls, then Romney should win convincingly. I suppose you could just conclude that the polls are crap.

Comments (6)

jussayin'
So much are defined by what they are trying to convey, and who commissioned the polls. There are a number of polls that were showing the gap narrowing between Romney and Obama (nation wide). Then the internal numbers showed a +7 Democratic sampling. There were others that had Obama over Romney by a couple of % yet the error rate was 7%! Yeah, that is a real good poll. Some of these polls are determined to push a narrative. What makes me think too many are full of crap is Romney has a large % of independents. An even % or close to it for women, a good deal +% of men, yet the overall result says its close. How? If they are nearly even with women, and Romney leads with men by a double digits, when last I knew those were the only two choices for humans, how can they be close? Many of the polls keep the internal numbers hidden. Why? Some samplings give 40% Democrat, 28% Republican. Yet national polls on party identification show an edge toward Republican. For too long polls have been used to drive the news. With less cooperation with the pollsters, how are we to trust these numbers?
Boyd
Every polling service uses its own methodology. Even RCP's "average of the polls" is pretty tenuous. Add to that the fact that many of the state-level polls are conducted by different organizations from the nationwide polls, and you've got...utter crap. I think you're trying to push the data somewhere it doesn't belong, CC.
CosmicConservative
Perhaps. But I think the statistical approach I am using is still valid. Adding up all the individual state polls gives a much, much larger sample size which ought to provide a lower margin of error. But what I have not factored in is the individual +D vs +R breakup of the samples. What this indicates to me is that if the national polls are accurate, then the state polls must be heavily skewed towards Democrats.
Drax
Latest polls (Gallup and Rasmussen) show that, nationally, self-identified Republicans slightly (1%) outnumber self-identified Democrats. That is HUGE. All the polls have a D+ (something - from 4-8 usually) sample. Plugging in the difference, you get about a 1-2% change in results. Then, we see that independents are trending HEAVILY toward Republicans. (Somewhere between 8-14%). Another HUGE fact. Then we have the fact that, in early voting, more Republicans turned out in Ohio than in 2008 and less Democrats did. All in all, I believe that all of these things are skewing the polling in Obama's favor. Accordingly, I'm with Dick Morris and others who believe that it'll be a 5% Romney win in popular vote which translates into a 320+ electoral victory.
Drax
Some of the above is supported in the following link, but it also shows that, nationally, Romney actually has a LEAD in early voting. (It also shows a tie in those who will vote early but have not yet, while Romney has an even larger lead in those who will vote on election day.) http://www.breitbart.com/Big-Government/2012/10/29/Gallup-Shock-Romney-Up-7-with-early-voters
CosmicConservative
Drax, I had read that story too. I am starting to get marginally more optimistic, but I am still waiting for the "October Surprise" I am sure will come. Plus I'm still waiting for the inevitable Romney fallback from the surge after the debates. I've never seen an election where a surge like that (what was it? 5-7 points on average?) did not have a rebound in the other direction, usually at least half of that.