Quickly, read Dick Morris on the subject. His comments are spot on, here. 
I will quote the whole of the article, here because it doesn’t break out easily.
Republicans are getting depressed under an avalanche of polling suggesting that an Obama victory is in the offing. They, in fact, suggest no such thing! Here’s why:
1. All of the polling out there uses some variant of the 2008 election turnout as its model for weighting respondents and this overstates the Democratic vote by a huge margin.
In English, this means that when you do a poll you ask people if they are likely to vote. But any telephone survey always has too few blacks, Latinos, and young people and too many elderly in its sample. That’s because some don’t have landlines or are rarely at home or don’t speak English well enough to be interviewed or don’t have time to talk. Elderly are overstated because they tend to be home and to have time. So you need to increase the weight given to interviews with young people, blacks and Latinos and count those with seniors a bit less.Normally, this task is not difficult. Over the years, the black, Latino, young, and elderly proportion of the electorate has been fairly constant from election to election, except for a gradual increase in the Hispanic vote. You just need to look back at the last election to weight your polling numbers for this one.
But 2008 was no ordinary election. Blacks, for example, usually cast only 11% of the vote, but, in 2008, they made up 14% of the vote. Latinos increased their share of the vote by 1.5% and college kids almost doubled their vote share. Almost all pollsters are using the 2008 turnout models in weighting their samples. Rasmussen, more accurately, uses a mixture of 2008 and 2004 turnouts in determining his sample. That’s why his data usually is better for Romney.
But polling indicates a widespread lack of enthusiasm among Obama’s core demographic support due to high unemployment, disappointment with his policies and performance, and the lack of novelty in voting for a black candidate now that he has already served as president.
If you adjust virtually any of the published polls to reflect the 2004 vote, not the 2008 vote, they show the race either tied or Romney ahead, a view much closer to reality.
2. Almost all of the published polls show Obama getting less than 50% of the vote and less than 50% job approval. A majority of the voters either support Romney or are undecided in almost every poll.
But the fact is that the undecided vote always goes against the incumbent. In 1980 (the last time an incumbent Democrat was beaten), for example, the Gallup Poll of October 27th had Carter ahead by 45-39. Their survey on November 2nd showed Reagan catching up and leading by three points. In the actual voting, the Republican won by nine. The undecided vote broke sharply â€” and unanimously â€” for the challenger.
An undecided voter has really decided not to back the incumbent. He just won’t focus on the race until later in the game.
So, when the published poll shows Obama ahead by, say, 48-45, he’s really probably losing by 52-48!
Add these two factors together and the polls that are out there are all misleading. Any professional pollster (those consultants hired by candidates not by media outlets) would publish two findings for each poll â€” one using 2004 turnout modeling and the other using 2008 modeling. This would indicate just how dependent on an unusually high turnout of his base the Obama camp really is.
And since these are professionals, they know this. They KNOW the results are not accurate.
It doesn’t take a genius to figure out why the polls are so biased. They want to suppress the GOP vote. As Guy Benson  explains:
Public polling is supposed to predict and reflect public opinion, not drive it. But by the looks of two consecutive national surveys, it seems as though certain media organizations are far more interested in achieving the latter end than the former. Yesterday, the Washington Postand ABC News published a poll purporting to show the presidential race tied at 47 percent.
As Dan mentioned , the poll ‘s partisan sample was a D+9, with a D/R/I of 33/24/36. This is preposterous. That would mean that this fall’s electorate will be two points more Democratic than the Democrat wave year of 2008. As a point of reference, the 2010 midterms showed Democrats and Republicans represented exactly evenly. In spite of this terrible sample (for which WaPo polls are becoming infamous ), the race is all knotted up. One crucial note from the internals: Romney is beating Obama among independents  by 14 points. Let’s be frank — if Mitt Romney wins indies by anything close to 14 points in November, Barack Obama will be a one term president. It’s that simple. But one risible poll wasn’t quite enough for this week apparently; Reuters has gotten in on the action  as well. Their samples have been notoriously bad  all cycle, too, and this latest survey is no exception. Like its WaPo counterpart, Reuters’ polling outfit concocted am identical, ludicrous partisan sample of D+9 (among adults, D+5 among registered voters). The new poll’s findings?
(1) Barack Obama “leads” Mitt Romney by 6 points overall, with independents woefully under-represented.
(2) Obama’s job approval rating is 48/47, a +1 result in a D+9 sample. That’s bad news for the president.
(3) Worse news for The One: Of the few independents sampled, only 41 percent approve of his job performance. As Jay Cost notes , that’s worse than Dukakis territory.
(4) The president is 10 points underwater on the economy — again, with a sample skew that should really goose his numbers on every question.
(5) Almost inexplicably, Republicans hold a four-point advantage on the survey’s generic Congressional ballot. I repeat, D+9 sample.
On item (3), I’ll direct you back to my point about the WaPo survey: If Obama drops that group by a somewhat healthy margin — let alone 41-59 — he loses. Period.
Any questions, class?