After an election day
that started with elation and ended with frustration, Mary Beth
Cahill, John Kerry's campaign manager, said: "The exit polls were
all wrong." They weren't all wrong, but neither were they all right
and neither were they the entire source of Ms Cahill's frustration.
Exit polls gather information from voters leaving the polls.
Poll takers ask people how they voted and also about who they
are and why they voted the way they did.
Television networks started using exit polls in the 60s to help
them predict who would win an election before the final votes
were in. Over most of that time the exit polls conducted by a
company called Voting News Service (VNS), were part of a system
that worked pretty well.
Problems with VNS were starting to show up before 2000. But that
year the system simply came apart and sprayed egg all over the
faces, clothes and studios of the television networks.
The networks decided to blow the VNS system all to smithereens
and then make it all over new. So they got together with the Associated
Press and formed an organization called the National Election
Pool. The owners would get poll results and would also sell those
results to other news organizations.
Edison Media Research and Mitofsky International put the system
together and conducted exit polls during the 2004 primaries. By
election eve everyone was confident that the system would work
just fine.
On election morning, November 2, exit pollers went to the polls
along with a record turnout of voters. Soon the data from those
first exit polls was released to National Election Pool owners
and subscribers. Then the fun began.
The data from an early exit poll is a bit like the score from
the early innings of a baseball game. The data can help you tell
who's playing well, but it's not enough, by itself to predict
the outcome of the game.
Not everyone votes in the morning. Many businesspeople skip long
lines at the polling place in the morning because they need to
get to work on time. They'll wait in line in the evening when
there's less time pressure.
Data is not enough. You need a prediction model. Those models
can only be based history and, historically, in most precincts,
increased turnout has correlated with an increase in overall votes
for the Democratic candidates. This time around, though, lots
of the new voters streaming to the polls and standing in those
long lines were voting Republican.
Then there were the Weblogs or blogs. On election day, their
reporting of early exit poll data wound up adding to the confusion.
Within minutes of the release of the first exit poll data, results
had been leaked to several blogs. Within minutes after that they
were on the net.
Some of the blogs made a weak attempt to tell viewers that they
were seeing early data. But none bothered to explain why early
exit poll results, by themselves, could be misleading.
The networks had paid for the data the blogs were reporting,
but they couldn't report it themselves because of their contracts
with the National Election Pool. But even though they couldn't
discuss the exit poll data directly, it clearly affected their
reporting.
In the early afternoon, commentators favoring the Democrats wore
big smiles and tended to chortle. Commentators favoring the Republicans
looked glum. But as the day wore on, those expressions changed.
It was almost like the commentators switched masks.
By the wee small hours of Wednesday morning it seemed clear that
George Bush had won Ohio and the Presidency. The Kerry campaign
did not concede defeat until mid-morning, no doubt taking time
to make sure that the numbers they were seeing were valid representation
of voter intent.
So, at the end of the day, did the exit polls get it all wrong?
No. By the time all the exit poll data was in the results were
within normal margins of error. Most of the problems involving
exit polls came from the way the data was published and used.
The raised expectations and dashed hopes of the 2004 election
started with the blogs. They reported raw data without review
or explanation. That raw data gave the conspiracy theorists and
wishful thinkers plenty to chew on. And that's why the mainstream
media, with their experts and fact checkers and editors won't
be going away any time soon.
Top of page
The
National Election Pool
A comprehensive Web site run by the University of Virginia's
Center for Politics, Larry
J. Sabato's Crystal Ball features analyses of presidential
elections, Senate, House and gubernatorial races.
Some blogs who carried exit poll data on Election Day 2004
Professional Groups that Set Standards for Polling
November 9, 2004
Reprinting and Reposting This Column
You may reprint or repost this article providing
that the following conditions are met:
- The article remains essentially unaltered.
- Wally Bock is shown as the author.
- The notice Copyright 2004 by Wally Bock or similar appears
on the article.
- Contact information for Wally is included with the article.
You may refer readers to this Web site as a way to meet this
requirement. Please link to http://www.bockinfo.com/
- Here is the wording we suggest when linking to this site.
"The article you've just read can be found on Wally Bock's extensive
Resource Web site along with many other articles and resources."
Any other reprinting or reposting requires specific permission
which is almost always granted. Click
here to request permission if necessary.
