We approach electoral integrity with a nonpartisan goal of transparency and honest and accurate recording and tabulation. But there is nothing nonpartisan about the patterns we keep finding. And there is nothing democratic about the veils that are being progressively lowered over what has traditionally been public information.
The public is clearly being shut out of the process. Where are the pollsters in all this? They're in a bind, aren't they?
I'm not quite sure what you mean by "a bind," but quite obviously a pollster's success is dependent on one thing--getting races right. As a pollster, you can have a dozen Certificates of Methodological Purity hanging on your wall but that won't keep you in business very long if you keep getting races wrong, especially if it's always in the same direction.
So, pollsters are in a bind in the sense that they have to do something to accommodate to the reality that competitive races keep turning out more Republican (or, in the case of primaries, more in favor of the candidate(s) that the radical Right would like to see win) than their polls, using best methodological practices, would predict. Enter the LVCM (Likely Voter Cutoff Model), pioneered a few years ago by Gallup (not the original George Gallup but an extreme right-wing descendant who now runs the operation). What the LVCM does is submit each potential poll respondent to a battery of (usually) seven questions, relating to such things as stability of residence, past voting participation, etc. It is known and accepted that the further to the right a voter is on the political spectrum, the more likely he or she is to answer these questions in the affirmative.
What the LVCM then does is establish a qualifying threshold (often these days set at seven out of seven, for maximum effect) of some number of these questions answered in the affirmative, and anyone below that threshold is excluded from the sample--that is, assigned a zero likelihood of voting. So disproportionately more of the left or Democratic part of the sample is excluded from participation. Result: an oversampling of Republican and right-wing voters. It would be methodologically justifiable to assign a lower weight to the responses of voters who answered fewer questions in the affirmative. But many of these respondents will in fact vote (if their likelihood of voting is, say, 50% then half of them will vote). To the extent they are excluded entirely from these pre-election samples, the polls should deliver a red-shifted result, to the right of actual vote counts honestly tabulated. Instead, as pollsters like to celebrate, the LVCM adjustment helps them get races right! You'd think someone (besides us) would think of asking "Why?"
The LVCM is a tunable fudge factor, pure and simple. From the pollsters' standpoint (and I was an analyst at a political polling firm years ago, before the computers took over our vote counting), theirs is not to question why the LVCM works, but just to employ whatever sampling hocus-pocus is necessary to get races right. So we have an abominable and unjustifiable methodology being used to match vote counts that are presumptively regarded as sacrosanct. And then, of course, those manipulated vote counts nicely jibe with the expectations established by the polls everything looks just fine. It's nicely circular.
You can often see the LVCM kick in a couple of months before the election, when the pollsters start shifting from a "registered voters" (RV) sample to a "likely voters" (LV) sample and the GOP gets a big, sudden "bump" that bears no relation to actual political circumstances. It happened this year with Gallup's Generic Congressional Ballot ("Who are you planning to vote for in November, the Democrat or Republican running in your district?"), and everyone (except a few of us lonely forensics types keeping honest score at home) just assumed America was veering right.
There's also this: all the polls, both tracking and exit polls, are weighted to the pollsters' best guess of the demographic composition of the electorate, and this weighting has an enormous impact upon the poll results. Where do they get these crucial demographic numbers? Generally from exit poll demographics established in prior elections. We know that since 2004, exit polls have consistently had to be "adjusted" to the right in order to match the vote counts. That is, the "Who did you vote for?" question is used to reweight the poll responses (it's called, with no deliberate attempt at irony, "forcing") to match the vote counts, and in the process every question on the questionnaire--including, e.g., "Are you a Democrat or Republican?" "What is your race?"--is reweighted as well. So, for example, when Edison/Mitofsky (the exit polling firm) needed more Bush voters to match the red-shifted vote counts in 2004 and adjusted their polls accordingly, Party ID got dragged along for the ride, and wound up at 37%D - 37%R, an obvious distortion when every other measure of the electorate showed at minimum a 39%D - 35%R Democratic advantage.
But the pollsters use that adjusted (and it's always rightward-adjusted) and distorted data to weight their own samples. Result: a further push to the right. When you combine this insidious effect with that of the LVCM, you wind up with polling results shifted anywhere from 5% to 10% to the right, a huge distortion and just about enough to cover all traces of outcome-determinative rigging of competitive contests in the upcoming election.
So, the pollsters are in a bind, sure; but polling, and especially sampling and weighting, are as much art as science, and pollsters have the flexibility to do what is necessary to "successfully" predict red-shifted (i.e., rigged) election results. The pollsters will stay fat and happy as long as they keep "getting it right," and they will keep "getting it right" as long as elections continue to be manipulated to the right and as long as the pollsters continue to use the LVCM and right-shifted demographic data to mirror the manipulation.
One little aside: at a conference in 2007, the lead pollster for the Democratic Party stated that when their own internal polling showed their (Democratic) candidate up by 10%, they had learned to regard the race as a tossup. Bear in mind that these are internal polls, designed not for the momentum-boosting effect of polls made public but to inform critical internal tactical decisions such as where to make last-minute media buys. The premium is on dead-on accuracy. Asked why he thought these polls were consistently "off" by 10%, he said he didn't know; he flat out discounted the suggestion that anything could be amiss with the vote counts. The never-happen-here wall of denial is that high.
Okay. It sounds like the pollsters are trying to have it both ways in order to hang onto their livelihood. Why hasn't the public had access to the actual data so we could evaluate this for ourselves? Isn't that what used to happen?
Short answer: because we live in a sham democracy. The idea, in our taxidermic specimen of a democracy, is that you go down to vote every couple of years and that's enough; that makes it a democracy. But we've ceded control of virtually every aspect of that process to a combination of supposedly public administrators whose motto is "see no evil . . ." and private corporations who are obliged to divulge nothing about their operations. We The People are not supposed to be interested in what's in the sausage; we're just supposed to eat it.
While there has been greater access to certain information in the past, the need for that information was much less critical when the vote counting process (and ancillary processes such as registration--who is allowed to vote) was observable. Now that the vote counting process is entirely concealed and unobservable, the only forensic measures we have left are indirect --polling, exit polling, demographic measures. When these numbers began waving red flags, and forensics experts began analyzing the disparities and "making trouble," the system responded by withholding the numbers and privatizing more and more information. "If thy eye offends thee, pluck it out."
In E2004, for starters, there was a massive exit poll vs.
vote count disparity that just happened to be highly concentrated in the
critical swing states, such as Ohio. There was also a sham
"explanation" given for this glaring red flag: Bush voters were more
reluctant to respond to exit pollsters. Now these exit polls were specifically weighted to neutralize any such effects, and it was these weighted polls that showed the glaring, outcome-determinative disparities,
but that was conveniently overlooked. As was the strange reality that there was no Reluctant Bush Responder effect in states like Utah or
Delaware where the election wasn't close, just in places like Ohio and Florida,
and New Mexico, where it mattered.
All this inconvenient truth was plowed under because the reality was just too disturbing. What would have settled the issue would have been for the exit polling poobahs (the Edison/Mitofsky firm and the NEP media consortium that controlled the information) to have released the precinct-level data for independent analysis. This they would not do; there was a lot of specious argument about confidentiality, though none of the respondents were identified individually so there was in fact no confidentiality issue. Edison/Mitofsky, under intense pressure, finally showed the data to a hand-picked non-expert who, surprise! surprise! said "There's nothing there folks; everybody return to your homes and your local TV programming."
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
4
2
2
