One poll done by two very respectable sources this week showed a margin of error that was "3.27%". That's pretty funny. Polls are a lot like movies. There is comedy, action, love, and continuing drama. Everybody is talking about polls, and many are even offering their opinions on them. The problem there is, everybody knows exactly what an opinion is like... and everybody has one.
With only a few exceptions, the people with an opinion have motives. This is where things get murky. When emotions and personal desire comes into play, scientific method can and usually does get forgotten. People bend results to what they want them to be, which taints the whole pool. Being a public pollster is tough. You're essentially a public opinion odds maker. In the case of the presidential race, you're creating a line on how a 100,000,000 or more people will react. For a professional football game, Las Vegas odds makers have less than 250 people and odds to calculate, yet the Vegas casinos can be different in their offerings. They also enjoy the added benefit of seeing where the gambling public is placing their money. Someone placing their money where their mouth is, is way better than a partisan opinion. Still, the odds makers can be WAY off, in fact it happens all the time. Odds makers get seasons of time to re-calibrate their accuracy, pollsters get two opportunities to see how right they are: primaries and general elections They also have to contend with what they see other pollsters doing, and this plays into some houses methodology. It can very much be a "keeping up with the Jones'" scenario.
The methodology difference is starkly apparent in the discrepancies of the results. Many houses get very different numbers from each other, and the numbers fluctuate non stop. Ask yourself this question, "how many times have I seen someone REALLY change their opinion on a topic like this?" The reality is that this kind of fluctuation is terribly wrong. Cataclysm aside, both Romney and Obama will end up with 43% of the vote each due strictly to party turn out. This leaves a top possibility of 14% of slack. Of that 14% of voters, it is reasonable to assume that a min of 4% (2%+2%} is committed both ways, getting to 45% each. Now what?
Now the goal of Romney's campaign is to get 48.9% (4.89% of the total) of the remaining voters., totaling 49.89%. But that's less than 50% of the vote!!! He'll lose!!! Wrong. Populous states like NY, CA, and IL will contribute to large overages in the popular vote but are of no consequence in the electoral college.
About the voter change, it does happen, but not so spastically. For any voter that is undecided at this point, the change is coming glacially, or instantly. The glacial change won't be done till November, and the instant change can happen at any time. Currently, president Obama is losing the instant change to foreign policy fiascos and people who originally disregarded him as a redistributionist. Romney isn't losing ANY votes over his 47% comment. While there are factual errors, the general tone is correct, and it most likely is not offending anyone already voting for him. He could have said it much better, and if the campaign defines the intent, it will turn into a positive. Conversely, there is no positive outcome in any regard for a tortured and murdered ambassador.
The polls say a lot, but the communication is not one way. They don't think, feel, and breathe, but voters do. To communicate with the polls these pundits should step out of "the story of the day" and assume a more long term view.
Besides, if the polls did all the work, and they were clinically precise, strategists would have to find new jobs, and you would not be reading this.