By DemocracyRules
Key learnings from the elections in 2000 and 2004:
The polls did not accurately predict election results in either of those years. Both predicted Democrat wins with substantial majorities. The basic polling methodologies were horribly unscientific, as I have detailed in previous postings.
The final measure of a science-grade survey or poll is the ability to accurately predict the future. A a bad polling methodology is no better than chance at predicting election results.
The same horribly flawed polling methods from 2000 and 2004 were used again in 2008. Pollsters used the same types of improper questions, improper interviewers, improper sampling methods, improper weighting methods, inaccurate data analysis, and strong leftist reporting biases.
Furthermore, there were more problems this time.
(1) About 80% of respondents refused to participate in the polls. That's more than usual, and it's a sign of trouble. When participation rates decline, the chance that the sample is representative also declines. There is also a possibility of sampling bias, where McCain/Palin supporters didn't want to reveal their voting preferences to strangers.
(2) The results of different polls don't even agree with each other. I call this "wobble". The poll is unreliable. That is, if you ran the same poll again, drawing from the same population, the results would be different. Wobble happens when the methodology is flawed. If a poll is unreliable, then it's not accurate either. More precisely, it's not valid. Valid results require reliable results. When polls wobble, as they're doing by more than 10 percentage points, they aren't valid, or accurate, or predictive.
In 2000 and 2004, thousands of polls were done, and almost all were wrong. Many tens of millions of dollars were wasted on them, and vast amounts of pundit's writings were completely wrong. Huge amounts of human memory capacity was polluted with false information.
In 200 and 2004, almost all polls predicted a Gore or Kerry win, or a Gore or Kerry landslide. Therefore, these polls were NOT operating at a chance level. Strangely, those thousands of polls which predicted a Gore or Kerry win were actually quite reliable if you just reversed their prediction!
If the polls said Gore or Kerry, then it was quite accurate to assume Bush would win. If you reversed the "valence" this way, most of the thousands of polls done in 2000 and 2004 predicted much better than chance!
Therefore, believe no exit polls, no pundits predictions, no MSM reports of "record turnouts in Democrat areas", or anything else anyone claims. Claims that someone "stole" the election are frivolous unless they can be proved in court.
There is a time to predict, and then there is a time to collect the actual data. From now on, the only data that count are the election results.
Pro Patria
Comments