*Why it's not a free country anyway
*Examples where the united $tates has not respected majority rule before
*See our article refuting a certain line of approach to seeing Florida as a fraud
*See our article on Zogby as a case of how capitalism influences polling
*About Republican conspiracy theories of exit poll rigging
*Knowing what to ask for in the debate about exit poll stats
January 20, 2005
Edison Media Research and Mitofsky International released a statement on their exit polls that showed Kerry winning the elections until after tabulations from the vote count entered the calculations--and even after tabulations from the vote count went into the weighting. The statements boasted that no projections were incorrect, because the system did not offer any winners until after use of tabulated votes as opposed to poll data.
The report of the exit-pollsters cannot be taken seriously at face-value, because as is typical under capitalism it is a marketing ploy. To say that no projections were wrong is to sidestep the question on the people's minds. The question is how there came to a be a discrepancy between the exit polls and the final Bu$h vote tally.
Factually, speaking, the Edison Media Research and Mitofsky International admitted to a programming error that resulted in an early incorrect reporting of the gender composition of the vote. They also admitted to having to reweight three states according to a post-election change of plans on absentee votes. Neither of these admissions addresses what critics of the U.$. vote said.
In addition, the exit pollsters admitted that there was a problem in 26 states where the Kerry vote was higher than in the tabulations and 4 states where the Bush vote was higher than in the tabulations. The pollsters checked the optical scan versus other techniques of voting and found some indications that there was no difference. However, that is only one of the problems that critics alleged after the elections, and it was not one that MIM endorsed prior to the pollster report.
The report points to within precinct error and not other weighting issues or technical hardware/server error as the source of the discrepancy of interest. The blame is laid firmly at the door of statistical bias and not a weighting problem. However, the exit pollsters offer no convincing within precinct error explanation. The notion that Republicans would respond less this election is the magic bullet they need of course, but such is a speculation with no more factual basis than saying that a few states were rigged.
The other problems that the pollsters point to have even less likelihood of being real sources of the problem. They simply make good marketing to point to--training of staff and distance from the polling booth for example. The pollsters' own data released in tabular form points to no obvious non-response compared with the previous election. What the pollsters really need if they care is some proof that precincts that are Republican had poll workers more likely to push away exit pollsters and cause non-response. However, in truth, the report shows that Republican-leaning precincts had perfectly fine non-response rates.
In fact, the report only deepens concerns about the elections. Facile explanations offered by the president's advisers such as that the gender composition was wrong go by the wayside as the pollsters show themselves that their first, second and third calls showed nothing but increasing margins for Kerry, late into the night. The gender problem was completely gone by 7:30 pm of election day, and substantially reduced even before that.
The most convincing argument in the report against a rigging is the picture of fearful Republican voters refusing to respond to stranger-pollsters more than 50 feet from the poll, but we did not get a sense that that happened often enough to account for the discrepancy compared with previous elections.
One of the most suspicious parts of the report is on page 42 where it shows that as MIM said in previous articles, the bias in swing states was higher than non-swing states. That hardly undercuts an argument for fraud being the explanation instead of statistical bias. Even the argument about the parallels with the 1992 election are unconvincing. The pollsters show that when voter concern is high and there are swing states involved discrepancies are higher. Far be it for MIM to deny that a rigging might have occurred in 1992, just not enough to swing it for Bush Sr. Operatives would not try to fix safe Democrat states for the Republicans and that argument remains standing today.
One last point we will raise that is not too relevant but of interest to MIM. The exit pollsters admit they botched the Latino vote by giving too much weight to the Cuban vote in one context. It's something we have to watch out for when a survey is done on an oppressed nationality in a larger context. There seems to be much trouble with the Latino surveys.
The other arguments of the exit pollsters apply to all elections, not just 2004--interviewer training, date of hiring and age. Again, it makes good marketing to raise these points, but they are not presented in a way to be relevant to the question of why there was such a discrepancy in within precinct error in 2004.
We urge consumers to focus in on the question of whether the exit polls can verify the Bu$h election, not how future polls will improve or how the pollsters really deserve the dollars or excellent reputation they get. The pollsters can always change the subject or say their principal concern was not verification. That does mean we should accept that answer. It is likely that better answers are yet to come from academia as the data has gone to the public through two institutions, the Roper Center and the University of Michigan.