McCain Might Win Tomorrow...

McCain Might Win Tomorrow...

First of all, the last sample taken by Zogby on Halloween night shows McCain up one point on Obama. It's only 1000 people, but Dick Morris thinks that the ads taking Obama to task for his association with Rev. "God Damn America" Wright may be turning the tide:

It’s only a one night poll, but John Zogby reports that his Friday night survey shows McCain leading Obama by 48 to 47. It’s only a one night poll (as opposed to the usual three day moving average) but it is 1,000 interviews. It is also over Halloween night! But it is the first poll in three weeks that shows McCain leading. What’s up? We think that the advertisement being run by GOPTrust.com is having an effect. It is an ad featuring Rev. Jeremiah Wright decrying America and calling it “the USA of KKK” while Obama sat, deaf-mute in the congregation. By bringing the shocking reality of Rev. Wright back to America’s TV screens; GOPTrust.com is performing an important public service. It is just not credible that Obama sat in the congregation for twenty years, asked the Reverend to officiate at his wedding and to baptize his children, titled his book after one of his sermons, and did not know the kind of vile, anti-American hatred he was spewing.

I also mentioned recently that the margin of error in the latest poll published by Investors' Business Daily was larger than Obama's lead over McCain. IowaHawk points out here that margin of error is based on assumptions about your sample size and how random the sample is (of course).

But I would make even another point that isn't there: When pollsters take a survey, they ask for demographic information about those they have surveyed, and then use that information (race, age, income level, party affiliation) to weight the raw data so that it more accurately reflects a national sample. Therefore, if 58% of those who answered your survey were female, you adjust the final data to conform to the 51% of voters who are actually female. That's all fine and good, but what does it do to the margin of error if your raw data reflects a sample that very far from the actual voting population? It isn't just a matter of miscomputing the adjusted data; the margin of error for underrepresented groups will also be screwed up.

This is no time to be saying, "My vote against Obama won't matter." Get out and there and vote. You do not want this to turn into a revolution a year or two down the line against the fascism that the Democratic Party represents. It will not be pretty.