Strengthening Constitutional Self-Government

No Left Turns

Exit Poll Theories

Long talk with Karlyn Bowman, AEI’s poll watcher extraordinaire. She says the problem with the exit polls was twofold: the HateBush vote may have got out early (while the GOP vote was more evenly spread throughout the day), and skewed the early exit polls. The secone wave of exit polls that cam eout late in the day were more accurate (especially on Senate races). Second, Mitofsky and the other exit poll wizards have never used raw exit poll results to report on the composition of the electorate (women, men, etc), until the end of the day, when the results are rebalanced against historical norms for group participation. This was not done yesterday by anyone, and everyone bought into the raw and skewed early figures.

My summary thought is that this exit polling debacle made this election into a six hour equivalent of the Dewey-Truman race in 1948.

Discussions - 3 Comments

So, is it true? Did the big media pass on bogus data to Wonkette?

Observation on the Ohio exit polls as reported on In most of the categories of focus, the final exit polls are in line with the actual vote for the State of Ohio.
The Ohio vote for Bush was right at 51%

According to the exit polls, Bush’s numbers as calculated by category of focus are as follows:

by gender: Bush 51.0%
by party registration: Bush 50.8%
by race: Bush 50.8%
by age: Bush 50.9%
by income: Bush 50.3%
by education: Bush 51.1%
by income: Bush 50.3%
by union affiliation: Bush 51.2%
by ideology: Bush 51.5%
by evangelical identity: Bush 51.2%

My conclusion at this point: the exit poll numbers are very reliable. And so it went with my number crunching until I came to exit poll numbers for the category focused on church/religious identification and attendance.

by church attendance: Bush 49.9%
by another version of church attendance: Bush 46.9%
by religion and attendance: Bush 48.3%

Huh? Something must be wrong here. Why would the exit polls in every other category of focus bear out the final results except for the categories that take various measures of church attendance? So I crunched the numbers again. Same result. Then it dawned on me, we’re missing some information here, these numbers do not add up to 100%.

For all the other categories of focus such as age, race, income... the totals add up to between 99-101%, what you would expect due to rounding, but for these the totals add up to 95%, 92%, and 94% respectively.

Trying to reconcile this, I rolled all of the "missing" voters into the "weekly Protestant" subset and low and behold, Bush’s Ohio vote suddenly becomes 51.8%. While slightly higher than the actual vote, it is a very similar in variance to that of other categories of focus such as ideology with a 0.5% variance, or income with a 0.7% variance from the actual result.

I then looked at the national exit poll numbers and the same statistical anomaly is there too, a 95% total instead of 100% and Bush underperforming at 48.2%. To be sure some categories of focus will not add up to 100% like "reason for voting" which cannot account for every reason, but given that these categories have subsets that could account for everyone, they should add up to 100%. For instance, one has an "all others" category if you’re not Protestant, Catholic, Jewish... Where, and more importantly, who are the missing 5%, 60% of whom appear to be Bush voters? I have a speculation on who they are but that is not the point here.

If these are the same numbers that were being viewed by pundits and campaign personnel on election day and they do not account for the "missing" voters, one would rightly conclude that Bush was significantly underperforming in the GOTV of one of his key constituencies, frequent Protestant church attendees, and if so, the election is over, Kerry is the next President.

Conspiracy? Probably not as in another category of focus used by the exit poll, "evangelical identity," the results are in line with the final results with only a 0.2% variance. A more likely explanation is either an error in the data entry grid that they used to tabulate the results or an error in instruction to poll takers as to how to indicate certain voters who did not fit the profile categories and they were thereby not accounted for thus skewing Bush’s exit poll numbers downward in a very key and closely watched constituency.

Just a thought.

"The hate-Bush voters must have gotten out early..."

One thing which distinguishes the red states from the blue ones is that people in the former, e.g., Southerners, are more civil than people in the latter. I believe there were a lot of hate-Kerry voters in this election, often people who seldom speak ill of anyone. He was giving aid and comfort to the enemy while our prisoners of war were being tortured.. Isn’t that enough? You see, I am one of those civil people, too, and feel the remainder of the catalog against him can be forgotten at this point. But there were extra millions of us in middle America who made damn sure we voted this time. There were something like nine million first-time voters over the age of 40. And we out-numbered the Bush-haters, didn’t we?

Leave a Comment

* denotes a required field

No TrackBacks
TrackBack URL:

Warning: include(/srv/users/prod-php-nltashbrook/apps/prod-php-nltashbrook/public/sd/nlt-blog/_includes/promo-main.php): failed to open stream: No such file or directory in /srv/users/prod-php-nltashbrook/apps/prod-php-nltashbrook/public/2004/11/exit-poll-theories.php on line 504

Warning: include(): Failed opening '/srv/users/prod-php-nltashbrook/apps/prod-php-nltashbrook/public/sd/nlt-blog/_includes/promo-main.php' for inclusion (include_path='.:/opt/sp/php7.2/lib/php') in /srv/users/prod-php-nltashbrook/apps/prod-php-nltashbrook/public/2004/11/exit-poll-theories.php on line 504