You're assuming that more than 29% of the country consider themselves Republicans.

Here's something to consider: There are Republicans, conservatives, paleo-conservatives, neo-conservatives, libertarians, Tea Partiers, birthers, deathers, and a few Constitution Party types.

On the other side there are just Democrats and the vague memory of the Green Party.

Any poll can get screwed up by not making their sample random enough. It happens. If you want to see if this particular poll is screwed up, and look at multiple other polls and see if they have similar results.

Not everything that's bad news for conservatives is a conspiracy.

The party affiliation is just as representative of the sample selected, as is the approval ratings. You could break it down to approval ratings based on party membership in the sample, but that tells you something totally different, and does not reflect a bias in sampling. That's why random selection polling is used; to reduce bias's in sample selection.

An analogy would be a fisherman trolling at sea and brings up his net. He has 46% Spanish Mackerel, 29% Red Snapper, 4% grouper, and 20% Rock Cod. That's your randomly selected sample. Then from that you can determine which food is most approved of in percentages. The food the Spanish Mackerel approves of most will probably get the highest approval rating, since there's more of them.

by party affiliation.

or else they'd be polling children and non-citizens as part of the sample as well.

Typically, political polls try for a "representative" random sample, which requires over-sampling certain demographics.

For example, they may want a proportionate number of blacks and latins, males and females, etc consistent with the overall population or of so-called "likely voters". Similarly, polls often have party representation consistent with registration rolls or last presidential vote shares.

It wouldn't tell us very much if by pure chance everyone in the sample ended up being a homosexual asian living below the poverty line. The odds of that occurring are practically nil, but even if not 100%, pollsters do understand that over-representation of certain groups skews results.

would you like a rhetorical answer?

Wow! The AP poll has Obama’s approval rating hitting 60 percent! And 53 percent say he deserves to be reelected.

And on the economy, 52 percent approve of the way Obama’s handling it, and only 47 percent disapprove! He’s up 54–46 on approval of how he’s handling health care! On unemployment, 52 percent approval, 47 percent disapproval! 57 percent approval on handling Libya! Even on the deficit, he’s at 47 percent approval, 52 percent disapproval!

BUT then you get to the party ID: 46 percent identify as Democrat or leaning Democrat, 29 percent identify as Republican or leaning Republican, 4 percent identify as purely independent leaning towards neither party, and 20 percent answered, “I don’t know.”

So the AP poll skews to Democrats by 17%.  Is this the first time they have ever conducted a poll or are they trying to convince the country that everybody loves Obama as much as they do?  Does the AP have a bias (GASP)?

And it was done by like 50% democrats, 30% independants and 20% republicans too...sounds like a nice, fair, poll, right!

Hate to have 48% democrats, 48% republicans and 4% independants, as the country really votes!  His approval ratings might even be double digits then!

are you forgetting that Clinton ran in a 3-way race both times?

In that case I'm sure we will find the AP using "random" sampling that skews 17% more conservative very soon. It's bound to randomly happen one of these days.  I'll just hold my breath and wait for that.

How about Clinton never getting 50% of the popular vote in two elections but polling 70% favorable?

Then what's the point of the poll?    What is the news value of a flawed, non-scientific poll?

about Palin, et al., you will be in the 90th percentile. You don't like the result so, you are suggesting the poll is no accurate or representative.

I like the skewing piece calculation by simple subtraction in complex statistical analysis. GREAT. In poll of this type, you want find out how much support you have within your party which you expect to be high, then you want to know which way independents are leaning next you want to see how much you might get from  the opposition.

It is well known fact that, neither party can elect their candidates by getting all the votes from party loyalists, all need votes from Independents.

More of a rhetorical question.

If every single Perot voter approved of Clinton it still wouldn't of equaled 70%.  And the Perot voters were dissatisfied Republicans.  But the myth was born that Clinton was overwhelmingly beloved with a 70% approval poll that is always cited when Clinton is discussed.  That's the purpose of these skewed polls.

...by that logic, Bush's 90% approval ratings after 9/11 was even a bigger myth.

Look, do you not understand the difference between voting for Candidate A, B, or C, and once the election is over, saying whether you approve or disapprove of the President's performance? You do understand that these are two seperate things, right?

Why not go to Harlem and ask how they think Obama is doing?

Or go to Salt Lake and ask them if Romney would be a good choice for President?

But of course, we would grab people at "Random"

You argument is TERRIBLY FLAWED!!