Will any mainstream pollster, and/or polling analyst–like Nate Silver–ever deign to talk about this matter?
From Jonathan Simon:
Stephen Herrington’s examination of the polling sleight-of-hand that occurs when pollsters move to “Likely Voter” samples as elections approach unfortunately misses a key point, perhaps the key point.
The “Likely Voter” samples, which so strongly favor the Republicans relative to the “Registered Voter” samples, are generated by the “Likely Voter Cutoff Model” (LVCM), first instituted several years ago by an extreme right-wing descendant of the reputable and venerable George Gallup.
What LVCM does is exclude (“cut off”) entirely from the sample any respondents who do not pass the seven-question “Likely Voter” test which Herrington reproduces and which is now a polling standard. Thus a whole group of voters who will in fact go to the polls (their aggregate likelihood of voting might be 30% or 50%) are assigned a zero likelihood of voting and dropped from the sample (a methodologically sound poll would weight responses based on respondents’ likelihood of voting, but not arbitrarily assign a zero weight, excluding them entirely). As Herrington notes, these excluded respondents are disproportionately Democratic voters. “Likely Voter” polls therefore substantially oversample Republicans and their results are skewed accordingly.
Here’s the rub: these Likely Voter polls are used and relied upon because, in the era of computerized voting, they keep getting important and competitive elections “right.” How can a poll that relies upon a methodological abomination “work” so well? No one–certainly not pollsters or the MSM–is bothering to ask this disturbing little question. Disturbing because the only rational answer is that the official vote-counts themselves are skewed Republican or “red-shifted.”
Election forensics experts have found the red-shift–rightward shift of vote-counts relative to exit polls, tracking polls, and hand counts–in every biennial election since 2002. What we’re seeing now, however, is that polling is catching up to the red shift. Tracking polls use the LVCM to account for the unexplained but pervasive pattern of competitive contests coming out more Republican than a methodologically sound poll would predict. And both tracking and exit polls are now weighted according to demographics (e.g., party ID) drawn from exit polls “adjusted” rightward to match red-shifted votecounts in prior elections, a further boost to Republicans.
So outcome determinative computerized manipulation of elections to the right now enjoys full cover from distorted tracking polls and exit polls. “Shocking” results are no longer shocking if they’ve been predicted by the polls. The LVCM is a big part of that story, since it adds to the weighting distortion derived from the “adjusted” exit polls of prior elections. It’s all sewn up rather neatly and, unless someone influential begins asking the disturbing little questions immediately, will ensure that election theft continues to determine the direction of America in this bizarre new world of computerized “democracy.”