http://www.nytimes.com/2000/11/09/politics/09SURV.html
November 9, 2000
THE POLLING ORGANIZATION
Data of Little-Known Service Led to Wrong Call in Florida
By NEIL A. LEWIS
Much of the seesaw confusion on Tuesday night over the presidential election
returns centered on the Voter News Service, a little-known but influential
consortium of television networks and The Associated Press that in recent
years has become the prime purveyor of voting results to the media.
The quick predictions by the major television networks around 8 p.m. that
Vice President Gore had won in Florida depended partly on calculations by
V.N.S., as the consortium is commonly called. Over the next two hours, the
networks, one after another, contritely withdrew their predictions and said
the state was too close to call. Many news organizations later awarded the
state to Gov. George W. Bush, again prematurely.
V.N.S. said in a statement tonight that it will investigate why models it
said had worked before "did not work properly in this situation." V.N.S.
said its survey of voters polls leaving the polls gave Mr. Gore a small lead
that was borne out by the early returns.
Most major newspapers, including The Times, who are not members of V.N.S.,
are subscribers to the service and use much of the same data.
V.N.S., the only service of its kind, relies on a combination of interviews
with voters as they leave the polls and actual vote counts to help its
members and other media organizations determine as soon as possible who is
winning an election. The organization is composed of CNN, Fox News, ABC,
CBS, NBC and The Associated Press. It is the latest result of the 20-year
evolution of media efforts, especially those of broadcasters, to declare the
winners as accurately and as early as possible.
Martin Plissner, the former executive political director of CBS News, said
that the agreement to set up a single cooperative system grew out of an
assessment by the networks after the 1988 election that it was too expensive
for each to do its own exit polling and predicting.
In 1980, NBC called the presidential election for Ronald Reagan at 8:15 p.m.
Eastern time, based on the networks's own surveys of voters leaving polls,
setting off an intense competition among the networks.
"So in 1984 all of the networks went in for huge exit polling operations"
and maintained those operations in 1988, Mr. Plissner recalled in an
interview. "One result was that the cost of calling elections, which up to
then had been relatively modest, suddenly became huge burdens to the news
departments of the networks."
So in 1990, they decided to create a joint operation.
At least one former network news executive said today that the deeply
embarrassing flawed calls about results in Florida this year were the result
of that cost cutting.
Tom Wolzien, the former executive producer of the "Nightly News" at NBC and
a former network vice president, said the decision to form V.N.S. was
short-sighted. "We're now seeing what happens when you bow to budget-cutting
pressures and end up with a single source that leaves you with no way to
compare and contrast different assessments."
Mr. Wolzien said that, for example, if the networks had been doing their own
polls and calculations as they once did, the public might have been given
one assessment that Mr. Gore won Florida, one that Mr. Bush won, and others
that said it was too close to call. He estimated that each network saved $5
million to $10 million this year by pooling their resources.
Mr. Wolzien, now a senior media analyst at Sanford C. Bernstein & Company,
said that the public is generally left to wonder why the networks all seem
to have the same predictions. While the networks and the subscribers to
V.N.S. have their own units to call races, they are all dealing with the
same data, he said. "That's why, last night, most calls were identical and
made within a few minutes of each other," he said.
Mr. Plissner, a former member of the V.N.S. board of managers, said that
exit polls, in which voters are asked how they voted as they leave the
polling places, were used until 1980 as a tool to analyze why elections turn
out the way they do, not to predict their outcomes.
The way V.N.S. works, he explained, is that a random scientific sample of
polling places is made. A V.N.S. worker who has been trained for a few weeks
beforehand stands near the polling place on Election Day, approaches voters,
and asks them to fill out questionnaires, which are then put in a closed
box. Typically there are three waves of exit polls, and as the day goes on,
those numbers are gradually replaced with actual results from each polling
place.
Those results are then analyzed according to statistical models created by
V.N.S., based on several factors, including past election performances.
David Pace, the projections editor at The Associated Press bureau in
Washington, said that he did not yet know what went wrong.
"The system is very good usually for getting accurate and early results," he
said. "But it's based on sampling techniques and the only time you can be
absolutely be sure is to wait to all the votes are counted. But we have such
a thirst to know, so it's based on the best science available."
Copyright 2000 The New York Times Company
