Replies

No. You're quite wrong. Rasmussen is not a Republican. He is non-partisan. I believe that after publishing a few Republican-leaning polls, and getting scads of attention from folks like Limbaugh, he began publishing what his audience, Republicans and Freepers, wanted to hear. He was a little nobody who found himself getting millions of snoopers, until he was shown to be dead wrong. Then he lost his ties to conservatives. His track record is lousy, and no-one would pay any attention for him, other then his automation gets results first.

Zogby is not a Democrat, either. He got big by doing polls for the New York Post, once it was taken over by the same company that runs Fox News. In 1993 and 1994, he was very influential in bringing Republicans to power in New Jersey and New York, such as Whitman, Pataki, and Giuliani, by publishing polls which showed those races as rapidly tightening. In fact, Murdoch was pushing Zogby's polls so hard, he cut the price of the POst in half, and then offered it free.

Then Zogby went national with Fox News, who dropped him.

Zogby always used innovative formulae for predicting races. And that's what his pools always were: poll-based predictions, not polls.

Stick to the pros: Gallup's surveys of REGISTERED voters. (Even Gallup has started corrupting their polls with prediction methods, hence, Gallup's "likely voters" polls. These polls are ridiculous. If you announce that Bush is up by 6 among likely voters, and down three among RVs, and liekly voters are 90% of your survey, what percentage does he get among the registered-but-not-likely voters? 10%? Fortunately, Gallup has money to spend, and does its best work in pre-sampling to make sure that its raw (RV) polls are excellent.)

Let me tell you all how a scientific poll would sample respondents. Rasmussen is mostly right, partly wrong.

Say about 75% of registered voters are party-affiliated likely voters, with a 39%D and 36% R, with a likelihood of voting of 85%. The other 25% are independents, with a probability of voting of 40%. To get the best results with the fewest calls, you want to under-sample the 75% and oversample the 25%. Why? What we're measuring here is sensitivity, changes, and only a few voters move from Party A to Party B, or vice versa, on a daily basis. But a voter may go from a 20% voter to an 80% voter..

Next, people lie. If you ask them how likely they are to vote, they will overstate. So you also look at actual data from previous elections. Note: these may be contaminated with vote fraud, probability approaching 100% in some precincts. E.g. city of Philadelphia reports 98% of all those over 18 registered. E.g. net vote fraud in 2000 Presidential election favored Gore by about 4,000,000 votes, 240,000 in Florida. However, therefore, polling should reflect expected vote fraud. Rasmussen in 2000 was new enough he didn't do it, and got disrespected for counting actual respondents without weighting for fraud.

Yes, once you use weighting, it's easy to manipulate the results. That's why the 2008 elections are important, to keep the 2010 census honest. We should take a tip from the Israelis of 2,010 years ago, and all live in rural areas (Idaho is best) on 1 April 2010. HHOS (ha-ha, only series).

And when you break the numbers down this way, it's still clear that turnout is key. Supposedly, conservative Christians didn't go to the polls in 2000 because of the DUI thing. I suspect that they did go, but in urban districts their votes went uncounted, thus explaining why more ballots are spoiled in urban than suburban precincts.

Finally, in Oregon and Washington, many to all votes are mailed in. Note how Senator Packwood was pushed to resign in Oregon just in time for vote-by-mail to take effect. Note that this is the method of voting that Stalin used in the 1920s to gain power. He said "It doesn't matter who votes. It only matters who counts the votes."

PS In the final analysis, the Second Amendment trumps the First.