woman sings "black national anthem" instead of star spangled banner.

Hear hear. That's just slightly more people than they poll for Celebrity Family Feud.

I can explain the statistics and math to you if you really want, but essentially, any poll of 1000 people - if sampled correctly and in an appropriately random fashion - is a more than adequate sample size to gauge the voting tendencies for an entire demographic.

I can explain the statistics and math to you if you really want, but essentially, any poll of 1000 people - if sampled correctly and in an appropriately random fashion - is a more than adequate sample size to gauge the voting tendencies for an entire demographic.Considering that more than 111,000,000 women (http://www.census.gov/population/socdemo/voting/cps2004/tab01.xls)(between the ages of 18-75) registered to vote in 2004, do you really think that you can an accurate random sample by only polling 1,000 people?

Even if you were only concerned about women who live with their spouse, you're still talking about a group of more than 59 million people.

Oh yeah, goes without saying that if you're trying to use a poll like that to draw conclusions about the larger electorate, you've got no idea what you're doing. So if that was MSNBC's intent, they're being retards, unless they're able to put it into context within overall state populations. But a poll like that does tell you how Obama is faring amongst a certain part of the population, which can indicate which messages he may push ahead with in the coming weeks/months. It also highlights the issues that are most important to certain parts of the population. For example, the working class and the poor are going to have a totally different set of priorities than the affluent.

You can also use these polls to examine historical data. Historically, there has been a very strong correlation between the candidate picked by certain demographic groups and the winning candidate. If you establish that historical correlation, you essentially establish that demographic as a swing vote. Polling that group can then be a helpful barometer for the upcoming election, provided you put the proper context around it.

I guess my real frustration has been how the media have sliced and diced the polls so much that it's all about race, gender, and class. And I don't like it. Yes, there is some redeemable value even in the most ridiculous poll I may disagree with. I just don't care for the larger narrative that the media tends to portray using this data - that is America is divided.

Considering that more than 111,000,000 women (http://www.census.gov/population/socdemo/voting/cps2004/tab01.xls)(between the ages of 18-75) registered to vote in 2004, do you really think that you can an accurate random sample by only polling 1,000 people?

Even if you were only concerned about women who live with their spouse, you're still talking about a group of more than 59 million people.

Yes. It's been mathematically proven. Go to any statistics professor at your locally respected university. He'll tell you the same thing.

I guess my real frustration has been how the media have sliced and diced the polls so much that it's all about race, gender, and class. And I don't like it. Yes, there is some redeemable value even in the most ridiculous poll I may disagree with. I just don't care for the larger narrative that the media tends to portray using this data - that is America is divided.

I'd agree that the media tends to use this information incorrectly. For one, I don't think the media has enough statistics experts to properly interpret data. Hardly any of them can explain what the margin of error really means.

But the bigger problem is they tend to push the notion that the nation is divided, as you say, because it fuels an "us vs them" mindset amongst the viewers. And that can be riveting, so they sell newspapers and ad space.

I think I'm on the same page as you. There are a ton of redeeming uses for breaking polls down by demographic. Unfortunately the media misses almost all of them.