News
Garber Announces Advisory Committee for Harvard Law School Dean Search
News
First Harvard Prize Book in Kosovo Established by Harvard Alumni
News
Ryan Murdock ’25 Remembered as Dedicated Advocate and Caring Friend
News
Harvard Faculty Appeal Temporary Suspensions From Widener Library
News
Man Who Managed Clients for High-End Cambridge Brothel Network Pleads Guilty
The five major television news networks have blamed inaccurate exit poll data for the series of false projections that left the country wondering who would be president and badly bruised media credibility.
After ABC, CBS, NBC, CNN and Fox News told viewers they believed a majority of Florida voters had chosen Al Gore '69 to be president, the state was put back in the 'too close to call' category.
And when a later call was made in Bush's favor, the networks were forced to rescind their predictions again.
The Voter News Service (VNS), a consortium of the five major news networks and the Associated Press, provides the exit polling data. On election days, it relies upon 40,000 poll-takers at key precincts throughout the country to paint a composite portrait of the nation's voting.
Statisticians and computer scientists work alongside demographers at VNS's New York headquarters, comparing exit polls to actual early returns. If they match to a degree--and if the match is verified by computer analysis--then VNS will project an outcome.
The moment-by-moment account of what happened Tuesday night and early Wednesday plays like a television drama.
At about 7:48 p.m., VNS sent a short coded message to the networks, whose own consultants, in turn, crunched the numbers.
At 7:50, NBC was first to declare Gore the winner in Florida.
Other networks did the same minutes later.
A Florida win would place an Electoral College win well within reach for Gore.
But then, a little before 10 p.m., VNS and the networks saw something strange: The actual tallies gave Bush the edge in the state, while the exit polls predicted a Gore win.
With disbelief in their voices, network anchors retracted the concession, placing the state back in their undecided columns.
"We don't just have egg in our face, we have omelet all over our suits," said NBC's Tom Brokaw.
For several hours thereafter, confusion reigned.
The Bush campaign called network political directors and insisted that key Republican precincts in the panhandle of Florida hadn't yet been counted, while Gore staffers claimed that the reliably Democratic counties of Metro-Dade and Broward in South Florida hadn't yet come in.
And then the networks made the same mistake again.
At 2:15 a.m, Fox News declared that Bush had, indeed, taken Florida.
That decision was based on actual votes, and not on data from VNS, according to a Fox spokesperson.
Within minutes the other networks followed suit, and Gore called Bush to offer his concession.
By this time, Florida vote counters had reported about 94 percent of the returns. Bush led by about 30,000 votes. As the percentages slowly grew, the margin of victory diminished rapidly. Those returns were largely from Democratic South Florida.
Suddenly, the margin was a mere 217 votes.
At 3:17 a.m., the networks began to pull their calls.
The next day, the networks blamed VNS, and VNS blamed statistical vagaries.
As analysts and pundits decry the networks' rush to judgment, it remains unclear precisely what caused the errors.
At 1 p.m. on Election Day, VNS exit poll data began to filter in to the news organizations.
Historically, exit polls have been accurate predictors of election outcomes, though some political scientists believe that as many as four precincts in ten may provide a false picture of the electorate.
Subscribers to the VNS service must agree to an embargo on the data until the majority of the polls in a particular state have closed, in order to ensure that a projected outcome does not influence voters or reduce turnout.
Networks have now accused VNS of providing bad data that led them to the make the multiple missed calls in the Florida race, and assert that the errors were not caused by poor internal analysis of that data.
"We based our call as did everybody else, on tabulated votes," said CBS news spokesperson Kim Akhtar. "As the data continued to come in after we made the call, we began to realize that something wasn't right."
Akhtar said that as the night wore on, the data was not as accurate as had been expected, and the actual numbers of votes didn't match up with VNS' predictions.
"Because of the problems in reporting results of the presidential race in Florida, CNN has initiated an immediate review of all procedures involved and has already begun consultation with the other news organizations," CNN said in a statement.
VNS is defending its data, though spokespersons have said that they will look into the source of the errors.
"These models, based on sampling precincts, have served us well through many elections. However, we will investigate why they did not work properly in this specific situation," VNS said in a statement.
Nonetheless, the pressure to be the first to project a state's returns appeared to come into conflict with industry standards of accuracy.
"We would rather be last than be wrong....If we say somebody's carried a state, you can pretty much take it to the bank," Dan Rather said before election coverage began on CBS last night.
Frank Newport, the president of Gallup organization, says VNS did the best it could under the circumstances.
"This is an extremely complex process done in 50 states and under intense pressure," Newport said. "It's amazing that they do as well as they do. It's not unusual that something will go wrong."
VNS did not return The Crimson's calls for comment yesterday.
Want to keep up with breaking news? Subscribe to our email newsletter.