Friday, April 01, 2005

TRUTH LEFT OUT: Data Indicates a Massive 2004 Exit Poll Whitewash

Image Hosted by ImageShack.us

Back in January, 2005, I emailed Dr. Steven Freeman about the Mitofsky anaylsis of the 2004 Presidential election exit poll. In that email, I made an observation about a particular detail in the Mitofsky analysis that had not been noticed by any researcher or critic of the report at that time. You may see the original observation I made here.

The Mitofsky hypothesis proposes that "within precinct error" (WPE) is the reason for the exit poll discrepancy that we've all heard about. In his report, he discusses many factors that he said contributed to this WPE. Factors such as: the distance the exit poll workers were from the actual election polling area; the weather conditions; interviewer characteristics (such as age) all contributed to the skew of the exit polls towards Kerry -- so the argument goes.

However, while I glanced over the data that Mitofsky used to justify this theory, something hit me over the head like a 2-by-4. In EVERY ONE of these factors that Mitofsky mentions in his report, THE WPE IS STILL SIGNIFICANTLY NEGATIVE IN THE MOST IDEAL CIRCUMSTANCE!!! That is, in Mitofsky's dream world, where the exit poll workers were:

1. working immediately adjacent to the ballot boxes.

2. interviewing electors where the weather was a nice sunny 70 degrees fahrenheit.

3. interviewing electors from only one precinct in multiple precinct areas.

4. between the ages of 55 and 64.

etc...

there still would be significant WPE, with the 2004 exit poll results.

As I stated in January, I was hoping that Freeman and others would do an analyses on this particular detail. Guess what? They did! In this latest report by Freeman et al., Analysis of the 2004 Presidential Election Exit Poll Discrepancies, US Vote Counts looked at all of the factors given by Mitofsky:

Image Hosted by ImageShack.us

First, we notice that even the lowest mean WPE errors for most factors are very high, even in the best of circumstances. We can therefore eliminate most of the above factors from consideration, leaving only population size of town or city, and voting equipment to consider.

For an example of why we can easily eliminate these factors, let us take "distance from polling place". The number of precincts where the pollsters were placed far from the polling station was small. The discrepancies that E/M seek to explain are already fully present even in the precincts where pollsters were optimally placed. Most of this bias is apparent even in the 75% of precincts where the pollster was allowed to conduct his survey just outside or within the building.

Hand counted paper ballots were used primarily in rural districts in only 3% of sampled precincts altogether, so had very little effect on the overall discrepancies. All voting methods produced higher mean WPEs in urban areas with over 50,000 population.

Rural areas constituted 24% of precincts sampled. All other "population size" precinct groups had mean WPE of at least -5.0, with the highest mean WPE of -7.9 in suburbs which constituted 39% of precincts.

No other factors relating to WPE (within precinct error) were given in the Edison/Mitofsky report that would explain the systematic discrepancies between the election results and the exit poll results in the presidential race.
So there you have it. I'll go even further, though. Since Mitofsky completely ignored this important detail -- and I refuse to believe the "father of exit polling" is incompetent -- one has to assume that Mitofsky is trying to explicitly MISLEAD us. For God sakes!!! I'm a simple layman who has no degree in statistics and I was able to spot this glaring detail!!!

Indeed, there is much more evidence offered by US Vote Counts that points to a 2 or perhaps 3-coat whitewash by Mitofsky.

According to this new study, there is absolutely NO EVIDENCE showing that Kerry supporters were oversampled. In fact, the evidence given by Mitofsky shows that the number of Bush supporters in a precinct is co-related to the exit poll discrepancy:

Image Hosted by ImageShack.us

The Mitofsky data also shows something which is highly improbable and counter-intuitive. Kerry supporters, according to the data, have a wide range of response rates, compared to the Bush supporters. What's even more improbable is that the Kerry supporters, according to Mitofsky's parameters, dramatically increase their response rate when they are in Republican territory. The same is not true for Republicans when in a Democratic precinct. As US Vote Counts muses:
This data contradicts previous experience and observations of this election that voters finding themselves in the minority in a local venue (and particularly a dwarfed minority) tend to be less willing to respond to exit poll interviewers, not more as this data requires. Certainly we would not expect the Kerry voter response rate to soar to over 84% in precincts where Bush voters outnumber them by at least four-to-one. Conversely, we would not expect the Kerry voter response rate to be at its lowest (53%) in precincts where Kerry voters were most numerous.
This is starting to look like an embarrassment to Mitofsky.

There are numerous different angles this new study analyzes with respect to participation vs. party affiliation. Time and time again, Mitofsky's data reveals that his hypothesis and assumptions are totally contradictory to normal polling behaviour. Here are the summaries of US Vote Counts observations on the data when related to Mitofsky's assumptions on respondent behaviour:
- Higher exit poll response rates and higher exit poll discrepancies occurred in Bush
strongholds. E/M’s own data contradict both the rBr and the rBrmpc hypotheses and support the Bsvcc hypothesis.

- The required pattern of exit poll participation by Kerry and Bush voters to satisfy the E/M exit poll data defies empirical experience and common sense under any assumed scenario.

- Once again, there is an implausible set of required response rates for Kerry and Bush supporters given the Edison/Mitofsky precinct partisanship data in Table 1.

- When Edison/Mitofsky’s explanation is checked against their own data using conservative assumptions, it requires highly suspect Kerry voter exit poll behaviors in "high-Bush" precincts.

- Very surprising patterns of partisan response rates to exit polls are required by both Kerry and Bush voters when we select values to minimize the differences between Bush and Kerry voters' exit poll response rates.

In sum, there are no values of proportions of Bush and Kerry voters which can be chosen that would result in plausible response rate patterns, and that satisfy the exit poll data given by E/M.
Get the picture?

Pretty dramatic stuff! Probably as dramatic as it gets for statistical analyses!!! But wait! There's more!!!

The exit polls for the 2004 election not only tabulated views from the Presidential election. It also received information about the voters intentions for the U.S. Senate races. Gues what?! Yup. Strangely enough, the exit polls were far more accurate at determining who would win for Senator. As history shows us, there is no precedent for widespread "ticket-splitting" in other elections. That is, if you vote democratic for President, there is an overwhelming probability that you would vote democratic for the Senator. US Vote Counts summarizes this peculularity this way:
There is no logic to account for non-responders or missed voters when discussing the
difference in the accuracy of results for the Senate versus the presidential races in the same exit poll.
No logic, indeed. Unless this is a nation where "multiple personality disorder" is present in epidemic proportions!!! To allay that particular fear, this report confirmed another startling finding which was observed in a previous report by the same group. Exit polling accuracy was dependent on whether the election ballots were hand-counted or not!! This is a highly significant finding, considering that, in Ohio, only a non-random 3% of the ballots were hand recounted. Many of these instances had recounts which were different from the machine counts. See this previous Newsclip Autopsy article that discusses this further.

In summary, the US Vote Counts report is damning in every conceivable way to the Mitofsky study. It has been shown that Mitofsky's conclusions completely contradict his own data. Furthermore, the only reasonable hypothesis that still has a leg to stand on is the one that supports the idea that widespread vote counting corruption occurred in the 2004 Presidential election. As US Vote Counts states:
An alternative hypothesis that is more consistent with the data is that corruption of the official vote count occurred most freely in districts that were overwhelmingly Bush strongholds.

If Edison/Mitofsky would release the detailed results of their poll to the public then much more could be said about this hypothesis, and the suspicious precincts could be identified. If E/M does not release its list of sampled precincts, US Count Votes believes it will still be possible to rigorously test the hypothesis that the vote counts were corrupted by assembling and analyzing a precinct-level nationwide database containing detailed election results, voting equipment information and demographic data.
Which brigs us to our last point.

When is the U.S. public going to finally get to see the raw exit poll data???!!!

Something tells me that this latest damning report is going to delay things a bit further. Especially when the MSM continues to refuse to fairly report what could be the biggest scandal in U.S. history.
Image Hosted by ImageShack.us
Name:
Location: Vancouver, British Columbia, Canada

(If you've got the urge - cut and paste)

info@ap.org; evening@cbsnews.com; comments@foxnews.com; letters@latimes.com; hardball@msnbc.com; nightly@nbc.com; newshour@pbs.org; letters@newsweek.com; nightline@abcnews.com; atc@npr.org; nytnews@nytimes.com; letters@time.com; tips@upi.com; editor@usatoday.com; letters@washpost.com; safire@nytimes.com; dabrooks@nytimes.com; fmanjoo@salon.com; mdaly@edit.nydailynews.com; nicholas@nytimes.com; andrewmsullivan@aol.com; rroeper@suntimes.com; suellentrop@slate.com; editor@weeklystandard.com; online@tnr.com; editor@watchblog.com; bobherb@nytimes.com; davidbroder@washpost.com; froomkin@washingtonpost.com; liberties@nytimes.com; public@nytimes.com; washington@nytimes.com; news-tips@nytimes.com; letters@slate.com; editpg@freepress.com; newsonline@bbc.co.uk;

Powered by Blogger

Weblog Commenting and Trackback by HaloScan.com
Listed on BlogShares